Dognosis Pairs Canine Brain Waves With AI to Detect Early Cancer

Unleashing Canine Superpowers - Cancer Screening That Doesn't Feel Like a Chore With Dognosis — Ai4

Ai4

Decoding canine neural responses via brain-computer interfaces removes human handler error to identify early-stage cancer at five times the sensitivity of liquid biopsy.

AI-assisted summary of Unleashing Canine Superpowers - Cancer Screening That Doesn't Feel Like a Chore With Dognosis — original source. Check the original for full context.

Source published: 2026-08-17

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 17, 2026, this presentation features Amir Alpert, Head of US Business Development at Dognosis. According to the video description, the talk explains how Dognosis combines trained detection dogs, brain-computer interfaces, and neural networks to decode canine olfaction into a clinically validated, highly sensitive at-home Stage I cancer screening platform.

## TL;DR Amir Alpert presents Dognosis, a diagnostic neuroAI platform that harnesses canine olfactory biology to detect multi-cancer biomarkers from a simple 10-minute at-home breath mask. The technology replaces subjective human dog handlers with an automated hardware and software stack—combining computer vision (FERAL), non-invasive EEG headsets (DogSense), and Bayesian integration (DogOS) to allow a 30-dog pack to analyze up to 1,000,000 patient samples annually. Backed by a landmark 3,300-participant trial in India demonstrating universal detection across 20 cancer types with 91.8% specificity, Dognosis plans US trials in 2027 and commercial at-home screening kits by 2028. Beyond oncology, the neural decoding platform is being leveraged for tactical law enforcement scent mapping and reinforcement learning frameworks to teach AI how to smell.

## Atmosphere & style The presentation blends high-tech Silicon Valley keynote enthusiasm with deeply human, historical, and biological storytelling. Alpert's delivery moves fluidly from sobering reflections on modern cancer statistics and healthcare costs to an affectionate, grounded on-stage presence alongside his dog, Snow. Watching the session feels both intellectually stimulating and warmly accessible, punctuating complex neuroAI architecture and EEG calibration telemetry with genuine optimism for the future of non-invasive preventative medicine.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=4s) Early-stage malignancies cause systemic metabolic shifts that release distinct volatile organic compounds (VOCs) into exhaled breath, which canine olfactory systems can detect. - [04:09](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=249s) Late-stage diagnoses account for a massive share of the United States' $250 billion annual cancer burden, underscoring the vital need for non-invasive early detection. - [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) Dognosis overcomes the historical bottleneck of canine scent detection by replacing human handlers with automated hardware, computer vision, and non-invasive neural telemetry. - [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) The platform's neuroAI stack (FERAL computer vision, DogSense EEG hardware, and DogOS Bayesian integration) allows a pack of 30 dogs to objectively process up to 1 million samples per year. - [11:29](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=689s) The patient screening experience is designed to be frictionless, requiring only 10 minutes of wearing a breath collection mask at home before mailing it to a centralized lab. - [11:29](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=689s) A clinical study across nearly 3,300 participants demonstrated detection across 20 distinct cancer types with 91.8% specificity, achieving roughly five times the sensitivity of liquid biopsy. - [14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) Canine olfactory brain decoding enables spatial scent data to be translated into real-time visual smell maps, as demonstrated in pilot operations with the Singapore Police Force. - [14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) Using canine neural responses as feedback mirrors Reinforcement Learning from Human Feedback (RLHF), creating a closed-loop framework to teach AI models how to smell. - [16:34](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=994s) Historical breast cancer advocacy reduced mortality by 50% through widespread public awareness and routine screening, serving as a model for scalable multi-cancer early detection. - [18:51](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1131s) Telemetry from the 'salami stimulus test' on the dog Snow confirms that canine olfactory recognition triggers immediate, quantifiable neural voltage spikes on non-invasive EEG headsets. - [20:14](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1214s) Dognosis combines cognitive science, military K-9 expertise, and neurotechnology partnerships with g.tec to shift AI's clinical focus toward accessible, non-invasive diagnostics.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=4s) The 30,000-Year Bond and the Biology of Canine Olfaction

Modern humans emerged roughly 300,000 years ago as a curious, migratory species—expanding into India 60,000 years ago, Europe 50,000 years ago, and across the Bering land bridge into the Americas 20,000 years ago. However, the foundational milestone for Dognosis occurred 30,000 years ago in modern-day Eurasia when humans first encountered wolves. This partnership went far beyond hunting larger game and guarding camps; nighttime protection from wolves is widely theorized to have allowed humans to sleep simultaneously and begin to dream. Over millennia, working wolves evolved into domestic companions, creating an emotional bond that roughly half of modern households still enjoy today.

```mermaid flowchart LR accTitle: Biological Pathway of Canine Breath Detection subgraph Biological ["Biological Mechanism"] A["Metabolic Shift & Immune Fight"] --> B["Volatile Organic Compounds (VOCs)"] B --> C["Bloodstream to Lungs"] C --> D["Exhaled in Breath Mask"] end subgraph Detection ["Canine Decoding"] D --> E["Canine Olfactory Receptors"] E --> F["Multimodal AI Signal Extraction"] end ```

Beyond companionship and protection, canines possess an extraordinary sensory capability: sensing systemic biological changes. When a person develops cancer, their entire metabolic profile shifts as the immune system responds, releasing volatile organic compounds (VOCs) into the bloodstream. These molecules travel directly to the lungs and are exhaled in the breath. For over 40 years, researchers have recognized that dogs can smell these signature VOCs. This was spotlighted in 2004 by the UK-based pioneering charity Medical Detection Dogs, whose founder's dog, Daisy, alerted to her early-stage breast cancer and sparked two decades of foundational canine olfaction research. Culturally, stories like *The Art of Racing in the Rain*—where the narrator dog Enzo senses his owner's wife's cancer but lacks the voice to warn her—have captured this tragic communication barrier.

- Humans first domesticated wolves approximately 30,000 years ago in Eurasia. - UK organization Medical Detection Dogs was established following Daisy's breast cancer alert in 2004, creating 21 years of foundational olfaction research. - Exhaled volatile organic compounds (VOCs) provide a non-invasive biological fingerprint of early-stage malignancy.

---

[04:09](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=249s) The Cancer Crisis: Early Detection and the Economic Burden

Cancer is not a recent byproduct of modern civilization; it predates *Homo sapiens*. The earliest fossilized evidence of malignancy is an osteosarcoma discovered in a 1.7-million-year-old hominin toe fossil from South Africa. Today, cancer represents a massive public health and economic crisis. In the United States alone, 2 million individuals are diagnosed each year. For the 1 million patients diagnosed at an early stage, modern interventions offer a viable path to remission and a second chance at life. For the other 1 million diagnosed at advanced stages, current therapies generally offer only marginal life extensions of several months to a year.

Addressing this burden costs the United States $250 billion annually. To put this sum in perspective, it represents ten times the annual budget of NASA, triple the entire budget of the U.S. Department of Education, and approximately $2,000 for every household in America—expenditures that are largely spent reactively rather than on early screening. While wolves became working partners 30,000 years ago and specialized breeds emerged over the last millennium, it is only within the last five years that breakthroughs in artificial intelligence, specialized neural architectures, and scalable compute have made it possible to translate canine olfactory superpowers into clinical multi-cancer early detection (MCED).

- **1.7 million years ago**: Earliest fossilized hominin cancer record (osteosarcoma toe bone in South Africa). - **2,000,000 US diagnoses annually**: 1,000,000 caught early with actionable therapeutic pathways; 1,000,000 diagnosed late with terminal prognoses. - **$250 billion annual US spend**: 10× NASA's budget, 3× the Department of Education's budget, or $2,000 per household.

---

[07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) The Dognosis Technology Stack: SniffSpace, FERAL, DogSense, and DogOS

The fundamental barrier to utilizing canine detection in clinical diagnostics has never been the dogs' olfactory acuity; it has been the communication bottleneck and lack of scalability. Traditionally, it requires over 12 months to train human handlers to interpret subtle canine behaviors, keeping the field confined to academic studies and anecdotal demonstrations. Dognosis overcomes this by completely removing human handlers from the diagnostic loop through an integrated hardware and software stack.

```mermaid flowchart TD accTitle: Dognosis Technical Platform Architecture subgraph Hardware ["Hardware Layer"] A["SniffSpace Bay (Visual & IR Cameras)"] B["DogSense BCI (Non-Invasive EEG)"] C["Sensor Vest (Respiration & HR)"] end subgraph AI ["AI & Translation Layer"] A --> D["FERAL Vision Model (V-JEPA 2)"] D -->|"Quality Sniff Window"| E["DogOS Bayesian Integration Engine"] B --> E C --> E end subgraph Output ["Clinical Diagnostic Output"] E --> F["Objective Sample Risk Score"] end ```

The system operates across three core pillars: 1. **SniffSpace & FERAL**: An automated sample presentation carousel and automated reward bay. Inside the bay, high-speed visual and infrared cameras feed into FERAL, a proprietary computer vision model built upon Meta's open-source V-JEPA 2. FERAL identifies the exact physical window of a "quality sniff" in space and time. 2. **DogSense Hardware**: While sniffing, dogs wear a non-invasive brain-computer interface (BCI) headset that records high-resolution EEG signals, alongside a sensor-equipped vest (the "doggy Iron Man suit") tracking respiration rates, heart rate, and fine micro-body language. 3. **DogOS**: A centralized Bayesian software platform that synchronizes the time-sliced FERAL sniff events with the corresponding EEG brain responses and autonomic physiological metrics, producing an objective, repeatable determination of sample danger without human interpretation.

Through this automated, multimodal neuroAI stack, a single pack of 30 trained dogs can process up to 1,000,000 patient samples per year.

---

[11:29](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=689s) Frictionless At-Home Screening and Landmark Clinical Validation

Patient adherence to cancer screening remains low because current procedures feel like an intrusive chore—requiring scheduling doctor visits, taking time off work, and undergoing uncomfortable procedures. Dognosis replaces this friction with an at-home breath collection kit. A patient simply wears a standard breath mask for 10 minutes at home while going about daily morning routines, such as reading the news or doing a word puzzle. The mask is placed back into the kit and mailed to a centralized Dognosis laboratory. If the canine AI analysis returns a high reactivity score, the patient is seamlessly referred into an insurance-reimbursable imaging workflow to pinpoint tumor localization and establish a definitive clinical treatment plan.

```mermaid flowchart LR accTitle: Patient Screening and Diagnostic Journey A["10-Min Home Breath Mask"] --> B["Mail to Centralized Lab"] B --> C["30-Dog Pack AI Evaluation"] C --> D{"Reactivity Score"} D -->|"High Risk"| E["Targeted Imaging & Physician Follow-up"] D -->|"Normal"| F["Annual Routine Rescreen"] ```

This model is backed by the largest canine olfaction clinical cancer trial conducted to date, run across nearly 3,300 participants in India in partnership with healthcare institutions including Narayana Health and Clearmedi Healthcare. The trial evaluated breath samples from patients across 20 distinct cancer types spanning seven major organ groups:

- **Universal Detection**: The platform demonstrated detection capability across all 20 tested cancer types without exception. - **Stage 1 to 4 Consistency**: High diagnostic sensitivity maintained across all stages (Stage I, II, III, and IV). - **91.8% Specificity**: Delivering a performance profile approximately five times more sensitive than conventional liquid biopsy assays at one-tenth the cost.

Dognosis is scaling its clinical validation with a randomized controlled trial of 30,000 participants (15,000 standard-of-care vs. 15,000 canine breath analysis). The company is establishing its US headquarters in Dallas, Texas, with formal US clinical validation trials scheduled for 2027 and commercial at-home kit availability targeted for 2028.

[14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) Canine Superpowers: Visual Smell Maps and Teaching AI to Smell

Beyond disease detection, canine olfaction unlocks two even larger technological frontiers. The second superpower is the ability to convert spatial olfactory data into visual maps. In an active pilot program with the Singapore Police Force, the team demonstrated how real-time canine scent tracking in the field can be translated into visual representations for police units, allowing officers to directly see what the dog detects in its environment. Human perception has long prioritized vision while remaining blind to the olfactory world; while GPS easily navigates people through Las Vegas to the nearest In-N-Out Burger, humans have no spatial awareness of the rich chemical landscape surrounding them. Reading canine brains brings this hidden dimension into clear view.

```mermaid flowchart LR accTitle: Canine Olfactory Decoding Pipeline

subgraph Sensing["1. Biological Sensing"] D["Detection Dog"] -->|"Olfactory Signal"| EEG["g.tec BCI / EEG Helmet"] end

subgraph AI["2. AI Processing"] EEG -->|"Neural Responses"| NN["Neural Network"] NN -->|"Reinforcement via Canine Feedback"| MODEL["Trained Olfactory AI Model"] end

subgraph Output["3. Real-World Applications"] MODEL -->|"Field Deployment"| VMAP["Visual Smell Maps (Police/Field)"] MODEL -->|"Clinical Pathway"| DIAG["Stage I Cancer Screening"] end ```

The third superpower introduces canine feedback into machine learning architectures, mirroring Reinforcement Learning from Human Feedback (RLHF) which propelled large language models from base systems to modern conversational agents like ChatGPT. By replacing human evaluators with canine neural responses, detection dogs validate correct and incorrect olfactory signals directly to the neural network. This closed-loop framework effectively teaches artificial intelligence how to smell.

- **Singapore Police Pilot**: Transforming raw olfactory detection data from working dogs into actionable, visual field maps for tactical law enforcement operations. - **Canine-in-the-Loop AI**: Applying the principles of RLHF to olfaction, using canine brain feedback to train models to accurately decode chemical signatures.

---

[16:34](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=994s) The 50-Year History of Breast Cancer Activism and Screening

The modern framework for cancer screening was not built overnight; it required more than half a century of relentless public advocacy, culture shifts, and destigmatization. In 1940, Maxine Davis took a monumental risk by publishing the groundbreaking article "Let There Be Light on Cancer" in *Good Housekeeping*, confronting a taboo disease that society refused to openly discuss. Decades later, in 1974, First Lady Betty Ford shattered political norms by publicly announcing her breast cancer diagnosis and recovery. That open declaration catalyzed widespread public dialogue, paving the way for Nancy Brinker to establish the Susan G. Komen Breast Cancer Foundation in 1982 following the death of her sister.

```mermaid flowchart TD accTitle: Timeline of Breast Cancer Screening Advocacy

A["1940: Maxine Davis publishes 'Let There Be Light on Cancer' in Good Housekeeping"] --> B["1974: First Lady Betty Ford publicly shares breast cancer diagnosis and survival"] B --> C["1982: Susan G. Komen Foundation established following Susan's passing"] C --> D["1992: Self Magazine launches the iconic Pink Ribbon campaign"] D --> E["Present Day: 50% decline in mortality since 1989 peak (~500,000+ lives saved)"] ```

The movement reached universal cultural visibility in 1992—52 years after Maxine Davis's article—when *Self* magazine introduced the iconic pink ribbon. That single symbol galvanized Breast Cancer Awareness Month and pushed mammography compliance rates far higher than any other cancer screening method. This collective activism yielded massive clinical results: since the peak in 1989, breast cancer mortality rates dropped by 50%, saving over half a million lives. Replicating this trajectory across other deadly, hard-to-detect cancers demands that same level of unified social activism and early detection awareness.

---

[18:51](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1131s) Meet Snow: Brain-Computer Interfaces and the Salami Stimulus Test

Dognosis operates an active campus in India consisting of 50 team members and an initial pack of 15 trained detection dogs. Expanding operations stateside, Amir brought his own dog, Snow—the inaugural member of the USA pack—directly onto the stage. While Snow was spared wearing the full BCI headset in front of the large conference audience, Amir shared live EEG recording traces captured earlier that morning during a calibration run dubbed the "salami stimulus test."

``` Raw Voltage (µV) ^ | /\ /\ /\ <- Salami Ingestion / Olfactory Peaks | / \ / \ / \ | /\ / \ /\ / \ /\ / \ ---+--/--\/------\----+--/--\/------\----+--/--\/------\----> Time (s) | Baseline Rest ```

The EEG telemetry revealed distinct, synchronized neural spikes corresponding precisely to the moment Snow caught the scent of salami. The biological mechanism remains identical whether a dog detects favorite food rewards or the microscopic volatile organic compound (VOC) biomarkers of early-stage tumors: intense olfactory recognition triggers immediate, measurable neurophysiological excitement. By pairing non-invasive brain-computer interface headsets with neural networks, AI can read and quantify those neural spikes instantly without relying solely on subtle canine behavioral cues.

- **Dognosis Pack Operations**: 50 human specialists and 15 trained dogs in India, alongside the newly established USA pack. - **The Salami Calibration Test**: Empirical demonstration showing immediate EEG neural response spikes upon olfactory target identification.

---

[20:14](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1214s) Driving Awareness: Early Detection, Clinical Partners, and the Dognosis Mission

Major clinical leaps against cancer do not solely depend on developing new chemotherapy compounds, experimental therapies, or viral discoveries on Reddit; they depend on early detection, broad awareness, and making screenings accessible before tumors advance. Having secured their seed funding round and validated the clinical assay, Dognosis is focused on user adoption and clinical deployment. Amir displayed a QR code on screen, inviting audience members to join the early registration waitlist and take a tangible stand for accessible, non-invasive cancer screening.

``` ┌────────────────────────┐ │ [Scan the QR Code] │ │ Early Waitlist Access │ └───────────┬────────────┘ │ ▼ ┌──────────────────────────────────────┐ │ • Join Early Screening Waitlist │ │ • Expand Clinical Advocacy Pool │ │ • Accelerate Scalable At-Home Tests │ └──────────────────────────────────────┘ ```

The venture originated as an unconventional collaboration between co-founders Ákos (a cognitive scientist from UC Berkeley) and Itamar (a former military K-9 specialist). What initially sounded like an impossible idea gained clinical momentum through partnerships with hospitals conducting rigorous clinical trials, neurotechnology hardware collaboration with g.tec medical engineering for canine EEG headsets, and documentary coverage led by David Savage at Harvey Nash Tech. Expanding the registry proves to healthcare systems and regulatory bodies that there is overwhelming demand for gentle, ultra-early cancer diagnostics.

> "The path to ending cancer is not about the next therapy. It is not about the next breakthrough that we see on Reddit. It is not about the next article, about the next chemotherapy or the next amazing medical improvement. It is about awareness."

- **Founding Roots**: Conceived by cognitive scientist Ákos (UC Berkeley) and military K-9 specialist Itamar. - **Key Technical & Media Partners**: g.tec medical engineering (BCI and EEG hardware integration) and David Savage / Harvey Nash Tech (documentary coverage). - **Mission**: Shifting the narrative of AI away from consumer apps and standard LLMs toward high-impact, life-saving neuroAI diagnostics.

## Notable quotes > What if the most sensitive Stage I cancer test on Earth isn't a lab assay — it's a dog's brain, decoded by AI? — Amir Alpert

> Dogs have been smelling cancer for decades; we taught AI to read what they smell. — Amir Alpert

> At Dognosis, we've built a clinically validated cancer screen by combining trained detection dogs, brain-computer interfaces, and a neural network that translates canine olfaction into digital diagnoses — Amir Alpert

> [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) The fundamental barrier to utilizing canine detection in clinical diagnostics has never been the dogs' olfactory acuity; it has been the communication bottleneck and lack of scalability. — Amir Alpert

> [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) doggy Iron Man suit — Amir Alpert

> [16:34](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=994s) Let There Be Light on Cancer — Maxine Davis

> [18:51](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1131s) salami stimulus test — Amir Alpert

> [20:14](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1214s) The path to ending cancer is not about the next therapy. It is not about the next breakthrough that we see on Reddit. It is not about the next article, about the next chemotherapy or the next amazing medical improvement. It is about awareness. — Amir Alpert

## Mentioned Amir Alpert, Dognosis, Medical Detection Dogs, Daisy, The Art of Racing in the Rain, Enzo, NASA, U.S. Department of Education, SniffSpace, FERAL, Meta, V-JEPA 2, DogSense, DogOS, Narayana Health, Clearmedi Healthcare, Singapore Police Force, In-N-Out Burger, ChatGPT, Maxine Davis, Good Housekeeping, Betty Ford, Nancy Brinker, Susan G. Komen Breast Cancer Foundation, Self Magazine, Snow, Ákos, UC Berkeley, Itamar, g.tec medical engineering, David Savage, Harvey Nash Tech

## Video description Amir Alpert - Head of US Business Development, Dognosis

What if the most sensitive Stage I cancer test on Earth isn't a lab assay — it's a dog's brain, decoded by AI? Dogs have been smelling cancer for decades; we taught AI to read what they smell. At Dognosis, we've built a clinically validated cancer screen by combining trained detection dogs, brain-computer interfaces, and a neural network that translates canine olfaction into digital diagnoses - 90.6% Stage I sensitivity, 91.3% specificity, peer-reviewed across 3,000 patient samples and roughly 5x more sensitive than liquid biopsy. The talk will walk through the technical stack (BCI hardware → EEG signal → neuroAI translation), the clinical pathway from canine signal to confirmed diagnosis, and the business model for at-home cancer screening at a price point that doesn't bankrupt patients. If you build AI for healthcare, work on multi-modal sensing, or want to see neuroAI deployed in the real world, this is the session.

[Source](https://www.youtube.com/watch?v=4FoUiWt8Cm8)

中文简报

Dognosis结合犬类脑电波与AI筛查早期癌症通过脑机接口解码犬类神经反应消除了训犬员的人为误差，能以五倍于液体活检的灵敏度识别早期癌症。 Amir Alpert介绍了神经AI诊断平台Dognosis，该平台利用犬类嗅觉生物学，通过患者在家佩戴10分钟的呼吸面罩即可检测多种癌症的生物标志物。该技术用自动化软硬件系统取代了主观的人类训犬员——结合计算机视觉（FERAL）、非侵入式EEG头戴设备（DogSense）和贝叶斯集成（DogOS），使一个由30只犬组成的犬群每年可分析多达100万份患者样本。在一项涵盖3300名受试者、证实其对20种癌症具有普遍检测能力且特异性达91.8%的印度里程碑式试验支持下，Dognosis计划于2027年在美国开展试验，并于2028年推出商业化居家筛查套件。除肿瘤学外，该神经解码平台还正被应用于战术执法的气味测绘，以及教导AI如何感知气味的强化学习框架中。早期癌症引发的代谢变化会释放特征性挥发性有机化合物（VOCs）进入呼出气体中，犬类敏锐的嗅觉系统能够有效捕捉这些生物指征。 (0:04) Dognosis 打造了由计算机视觉（FERAL）、非侵入式脑机接口（DogSense）和贝叶斯软件（DogOS）组成的神经 AI 技术栈，完全移除了传统训犬员的人工瓶颈，使 30 只工作犬每年可处理高达 100 万份样本。 (7:15) 在一项涵盖近 3,300 名受试者的临床试验中，该平台在 20 种不同癌症类型中实现了 91.8% 的特异性，灵敏度约为常规液体活检的 5 倍，且仅需患者在家佩戴呼气面罩 10 分钟后邮寄送检。 (11:29) 通过解码犬类嗅觉脑电信号，Dognosis 不仅在与新加坡警方的合作中实现了实时可视化气味地图，还借鉴 RLHF 原理以犬类神经反馈闭环训练 AI 模型学习识别人类未知的气味环境。 (14:58)

Reading summary

Dognosis combines canine olfaction with neurotechnology and artificial intelligence to create a scalable, non-invasive multi-cancer early detection platform. By reading neural and physiological signals directly from trained dogs, the company aims to eliminate the human-handler bottleneck and make at-home screening accessible. Biological Olfaction and the Early Detection Crisis (0:04) Cancer shifts metabolic activity and prompts immune responses that release volatile organic compounds (VOCs) into the bloodstream, which are subsequently exhaled through the breath. While researchers have recognized canine sensitivity to these signature VOCs for over four decades—spurred notably in 2004 by the UK charity Medical Detection Dogs—early clinical application has faced practical limits. Catching malignancies early remains a critical determinant of survival and medical costs. In the United States, roughly 2 million people receive cancer diagnoses each year; half are diagnosed late, leaving them with terminal prognoses and limited therapeutic options. Reactively treating advanced disease drives a $250 billion annual expenditure in the US, underscoring the economic and public health need for accessible early screening tools. Automating Olfactory Decoding with NeuroAI (7:15) The primary obstacle to clinical adoption of canine screening has been training human handlers to reliably interpret subtle canine alerts, a process that traditionally takes over a year per handler. Dognosis addresses this bottleneck by removing the human interpreter through an automated multimodal hardware and software architecture. Trained dogs enter an automated bay called SniffSpace, where infrared and high-speed cameras run FERAL, a vision model adapted from Meta's V-JEPA 2 to detect the precise window of a sniff. Simultaneously, a non-invasive EEG headset developed with g.tec medical engineering and a sensor vest capture the animal's neural responses, respiration, and heart rate. A central Bayesian engine, DogOS, synchronizes these signals to generate an objective risk score. According to the company, this setup allows a pack of 30 dogs to process up to 1 million samples annually. At-Home Screening and Landmark Clinical Validation (11:29) To lower friction for patient adherence, Dognosis utilizes an at-home breath collection kit requiring a user to wear a standard mask for 10 minutes. The sample is mailed to a central testing laboratory; if evaluated as high risk, the patient is referred for diagnostic imaging to locate the tumor and guide standard-of-care clinical management. In a trial involving nearly 3,300 participants in India alongside partners including Narayana Health and Clearmedi Healthcare, the platform demonstrated detection across all 20 tested cancer types, maintained sensitivity across stages I through IV, and achieved 91.8% specificity. Dognosis reports this performance exceeds liquid biopsy sensitivity fivefold at one-tenth the cost. The company is currently planning a 30,000-participant randomized controlled trial, with US clinical validation planned for 2027 and commercial launch targeted for 2028. Canine-in-the-Loop AI and Olfactory Mapping (14:58) Beyond diagnostic screening, the team is deploying canine sensory feedback to train artificial intelligence models to decode chemical environments. Mirroring Reinforcement Learning from Human Feedback (RLHF), canine neural responses validate correct or incorrect olfactory identifications, effectively creating a canine-in-the-loop training system for synthetic olfaction. This capability also extends into spatial mapping. In an ongoing pilot with the Singapore Police Force, tactical teams translate raw canine scent tracking into visual representations, allowing law enforcement officers to visualize chemical trails and environmental scent data in real time. Dognosis positions canine neuroAI as a potential solution to early cancer screening barriers, drawing inspiration from breast cancer advocacy campaigns that sharply lowered mortality. However, the technology's broader adoption remains dependent on the outcome of its planned 30,000-person randomized trial, upcoming US regulatory validation in 2027, and successful commercial scaling.

精炼阅读摘要

医疗科技公司 Dognosis 提出，通过将犬类敏锐的嗅觉能力与非侵入式脑机接口及人工智能相结合，能够消除传统训犬沟通瓶颈，实现低成本、无痛感的居家早期多癌症筛查。该团队认为，抗击癌症的关键在于提升超早期检测的可及性与公众意识，而非仅依赖晚期治疗药物。癌症经济负担与犬类呼气检测机制 (4:09) 癌症并非现代文明的副产物，在南非出土的170万年前古人类脚趾骨化石中便发现了骨肉瘤的痕迹。如今在美国，每年新增约200万例癌症确诊病例，其中早期确诊的100万人具备明确的缓解与康复路径，而晚期确诊的100万人通常只能获得数月至一年的有限生命延长。美国每年为此承担高达2500亿美元的医疗支出，达到美国国家航空航天局（NASA）年度预算的10倍，相当于平均每户家庭承担2000美元，且绝大部分支出耗费于被动的晚期治疗而非早期筛查。从生物学机制来看，当人体形成肿瘤时，代谢改变和免疫反应会产生特异性的挥发性有机化合物（VOCs）。这些分子随血液流入肺部并通过呼气排出。虽然科学界与相关慈善组织（如英国医疗侦测犬机构自2004年起）在过去数十年里已证实犬类能够敏锐嗅出这些早期病理气味标志物，但缺乏可扩展的工程化转化手段，长期限制了这一生物感知能力的临床普及。神经AI系统解决人犬沟通瓶颈 (7:15) 长期以来，利用犬类进行临床诊断的最大障碍并非其嗅觉灵敏度，而是训导员需要耗费12个月以上学习解读犬类的细微行为，导致测试效率低下且难以标准化。Dognosis 开发了一套完全排除人工驯导判断的技术架构，由自动化样本展示与奖励隔间（SniffSpace）、高分辨率脑机接口（DogSense）及中央贝叶斯软件引擎（DogOS）构成。在检测过程中，基于 Meta V-JEPA 2 架构的视觉模型 FERAL 通过高速红外与可见光摄像头锁定犬类嗅闻的精确有效时间窗口。同时，犬类佩戴由 g.tec 参与协作的无创脑机接口头盔与监测呼吸、心率的传感器背心，实时记录脑电图（EEG）信号。正如现场展示的“意大利香肠校准测试”所验证的，当犬类识别出目标气味时，其脑电遥测数据会立即出现高度同步的神经放电峰值。通过 DogOS 整合生理与脑电指标，一个由30只受训犬组成的系统每年能够客观、规模化地处理多达100万份病患样本。居家采样流程与大规模多癌种临床验证 (11:29) 针对传统癌症筛查因流程繁琐、侵入性强而导致依从性低下的问题，该方案采用无感化的居家呼气采样机制。受试者仅需在早晨进行日常活动时佩戴呼吸面罩10分钟，随后将其密封邮寄至中央实验室。一旦犬类AI系统评估出高反应性风险评分，患者将被转介至医疗保险覆盖的影像学检查流程以锁定肿瘤位置并制定针对性治疗计划。该系统在印度与 Narayana Health 及 Clearmedi Healthcare 合作开展了涵盖近3,300名受试者的临床试验，评估了横跨7个主要器官系统的20种不同癌症类型。试验数据显示，该平台对全部20种癌症均展现出检测能力，并在I期至IV期病程中保持一致的敏感性，特异性达到91.8%。团队指出，该性能的敏感度约为常规液体活检的5倍，而成本仅为其十分之一。目前团队正在推进包含30,000名受试者的随机对照试验，设立了达拉斯美国总部，计划于2027年开展美国临床试验，并目标在2028年实现商业化套件上市。嗅觉AI拓展与公共倡导的历史启示 (14:58) 除了疾病筛查，犬类嗅觉解码还在开拓新的技术边界。在与新加坡警察部队的合作试点中，工作犬的实时气味追踪数据被转化为执法人员肉眼可见的视觉地图，填补了人类对化学环境感知能力的空白。在机器学习架构层面，团队借鉴大语言模型的人类反馈强化学习（RLHF）机制，以犬类神经生理反馈取代人工标注，通过闭环网络训练AI识别和解析复杂的化学特征。演讲者强调，普及早期筛查需要复刻乳腺癌倡导运动的历史经验。从1940年破除禁忌的科普文章、1974年第一夫人贝蒂·福特公开患病，到1982年建立苏珊·科曼基金会及1992年粉红丝带符号的推广，半个世纪的公众倡导推动乳腺癌死亡率自1989年峰值以来下降了50%，挽救了超过50万条生命。团队认为，扭转癌症危机的根本不在于等待下一种化疗药物，而在于通过提升全社会对温和、超早期筛查技术的认知与采纳。 Dognosis 展示了一种结合生物嗅觉与神经AI的创新筛查路径，其技术核心在于用客观脑电与生理数据取代人工训导判断。不过，其实验室呼气筛查在全美临床的推广仍有待2027年验证试验与2028年商业化目标的落地检验，能否在真实世界中达到预期采纳率并与既有医疗保险及影像确诊体系顺畅衔接，仍是后续面临的关键挑战。

Detailed notes — zh

## Background 该视频由频道 Ai4 于 2026 年 8 月 17 日在 YouTube 平台发布，由 Dognosis 美国业务拓展负责人 Amir Alpert 带来主题演讲。视频内容围绕 Dognosis 如何将经过医疗训练的侦测犬、无创脑机接口（BCI）硬件与神经 AI 翻译算法相结合，把犬类对癌症气味的嗅觉感知转化为数字化客观诊断，从而打造一种高灵敏度、低成本且无痛的居家癌症早筛新范式。 ## TL;DR Dognosis 针对传统癌症筛查依从性低和成本高昂的痛点，构建了融合自动化气味仓（SniffSpace）、计算机视觉（Feral）、无创脑电传感（DogSense）与贝叶斯算法（DogOS）的多模态技术栈，彻底移除了人类驯犬员的主观沟通瓶颈。在印度开展的 3300 例临床试验中，该技术覆盖 20 种癌症并实现了 91.8% 的特异性，对 I 期癌症的敏感度达到传统手段的近 5 倍。团队不仅展示了将气味转化为战术视觉地图与训练“嗅觉大模型”的前沿成果，还公布了计划于 2028 年在美国推出 10 分钟居家呼吸早筛试剂盒的路线图。 ## Atmosphere & style 整场演讲充满科技乐观主义与深厚的人文情怀，将跨越三万年的人犬共生演化史、严谨前沿的神经工程学与抗癌公共卫生倡导紧密交织在一起。演讲风格生动而富有激情，当演讲者将爱犬 Snow 带上讲台并用当天实测的“萨拉米香肠脑电测试”展示神经电信号特征时，现场气氛在严谨硬核与轻松欢快之间达到了绝佳平衡，极具感召力与说服力。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=4s) 人犬共生可追溯至3万年前，犬类敏锐的感官不仅助力了人类早期演化，也为其现代医疗探测潜能奠定了生物学基础。 - [02:20](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=140s) 肿瘤代谢改变会产生特异性挥发性有机化合物（VOCs）并通过呼气排出，犬类嗅觉早在40年前就被发现能察觉此类异常，但此前受制于沟通与表达瓶颈。 - [04:49](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=289s) 美国每年抗癌医疗支出高达2500亿美元，但一半确诊患者已处于中晚期，早期发现与晚期确诊之间存在极大的生存与经济鸿沟。 - [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) Dognosis 构建了包含自动化气味仓 SniffSpace、V-JEPA 2 微调视觉模型 Feral、非侵入式脑机接口 DogSense 及贝叶斯算法 DogOS 的完整多模态技术栈，消除了人为主观解读误差。 - [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) 仅需一个由30只训练犬组成的检测团队，在自动化流水线下每年即可完成高达100万份临床呼吸样本的高通量检测。 - [11:29](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=689s) 在印度覆盖近3300例样本、20种癌症类型的临床试验中，系统实现91.8%的特异性，I期癌症敏感性达液体活检的近5倍，且检测成本仅为其十分之一。 - [11:29](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=689s) Dognosis 计划于2028年在美国推出居家无创呼吸早筛盒，将筛查过程简化为早晨佩戴10分钟面罩，解决传统早筛依从性极低的问题。 - [14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) 通过与新加坡警队合作，Dognosis 正在将警犬嗅探到的微量气味信号实时解码并转化为警员战术界面上的可视化空间地图。 - [14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) 借鉴大模型的 RLHF 范式，Dognosis 以犬类大脑电生理信号为真值标签构建嗅觉强化学习（RLDF），致力于教会神经 AI 大模型分辨数百万种挥发性分子。 - [17:34](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1054s) 回顾粉红丝带倡导运动表明，公众意识与普及性筛查是降低乳腺癌50%死亡率的关键，早筛革命需要强有力的社会认知推动。 - [18:51](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1131s) 现场展示的爱犬 Snow 脑电波测试验证了气味刺激会直接引发特征神经放电峰值，证明脑机接口与 AI 能精确数字化读取生物感知信号。 - [20:55](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1255s) 战胜癌症不能仅依赖下一代新型疗法，更需要借助 AI 与生物智能将普惠、无痛的早期筛查技术推向大众医疗体系。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=4s) 人类与犬类跨越三万年的共生奇迹 Dognosis 的核心理念在于坚信犬类拥有与生俱来的“超能力”，而团队的使命就是借助现代科技释放这种潜力以造福人类。回顾人类演化史，智人在约 30 万年前出现后不断迁徙探索：6 万年前进入印度，5 万年前踏足欧洲，2 万年前跨越白令海峡进入美洲。然而，演化史上最关键的奇迹发生在 3 万年前的欧亚大陆——人类首次与狼相遇。这段特殊的共生关系彻底改变了两个物种的命运。人类利用狼敏锐的感官和猎捕能力围捕体型更大的猎物；狼群则在夜间为人类营地提供警卫，甚至有学说认为正是狼的守护才让人类群体首次能够安心同时入睡并进入深度梦境。在数万年的演进中，狼被驯化为狗，成了人类不可或缺的工作伙伴与家庭伴侣。直至今日，犬类仍在从事狩猎、搜救和护卫工作，但它们的潜能远不止于此。 - **人类迁徙与驯化时间线**：30 万年前智人演化；6 万年前进入印度；5 万年前进入欧洲；3 万年前欧亚大陆狼与人类结盟；1000 年前人工选育现代犬种。 - **演化价值**：夜间警戒促进了人类集体睡眠与梦境的产生，建立了深厚的情感共生纽带。 --- ### [02:20](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=140s) 嗅觉与癌症的生物学信号：从历史、文学到演化化石当人体出现癌变时，全身的代谢机制会发生显著改变，免疫系统随之响应。这种代谢重塑会产生特定的挥发性有机化合物（VOCs），并随血液循环进入肺部，最终通过呼吸排出体外。科学界早在 40 年前就发现犬类能够凭借极其敏锐的嗅觉察觉到这些挥发性分子。英国先驱机构“医疗侦测犬”（Medical Detection Dogs）在 2004 年由创始人成立，起因正是其饲养的宠物犬 Daisy 提前嗅探并警示了主人患有乳腺癌。犬类感知疾病却受限于表达能力的情感张力，在文化作品中也屡有体现。例如小说《我在雨中等你》（*The Art of Racing in the Rain*）中，叙事犬 Enzo 很早就感知到女主人的身体出现恶化（乳腺癌），却因为无法用人类语言沟通而眼睁睁看着她离世。这揭示了犬类癌症嗅探的长期困境：犬类具备感知异常的能力，但缺乏精准、客观的沟通渠道向人类传达预警。与此同时，癌症的存在甚至早于人类文明——南非发现的一具 170 万年前的早期人类足趾化石中就已检测出骨肉瘤（Osteosarcoma），证明这种疾病自始至终伴随着演化进程。 - **生理机制**：肿瘤代谢改变 $\rightarrow$ 产生特异性挥发性有机化合物（VOCs） $\rightarrow$ 血液运输至肺部 $\rightarrow$ 经呼吸排出。 - **先驱案例**：英国医疗侦测犬组织（Medical Detection Dogs）及传奇侦测犬 Daisy。 - **古病理学证据**：南非 170 万年前人类足趾化石中已发现骨肉瘤病灶。 --- ### [04:49](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=289s) 癌症早筛的现实困境与两千五百亿美元的医疗支出在美国，每年有 200 万人被确诊患有癌症。其中 100 万人在早期被发现，能够通过现代疗法获得显著的生存改善和治愈机会；而另外 100 万人确诊时已处于中晚期，现代疗法往往只能将生命延长数月至一年，确诊几乎等同于判决死刑。早期发现与晚期确诊之间的生存鸿沟极其巨大。为抗击癌症，美国每年消耗高达 2500 亿美元的医疗资金。这一庞大数字相当于美国国家航空航天局（NASA）年度预算的 10 倍，是美国教育部总预算的 3 倍，折合每个美国家庭需分摊约 2000 美元。尽管投入巨大，现有的筛查手段依然难以大规模普及。犬类嗅探癌症的能力虽然被认知了 40 年，但直到近 5 年随着人工智能架构演进、计算能力提升与多模态传感技术的突破，人类才真正具备了将这种“嗅觉超能力”转化为标准化、规模化泛癌种早期筛查（MCED）工具的技术条件。 ```mermaid flowchart TD accTitle: 癌症筛查现状对比与早期发现价值 subgraph CurrentCrisis["每年全美 200 万癌症确诊患者"] A["100 万早期确诊患者"] --> B["现代疗法干预 / 治愈希望大"] C["100 万晚期确诊患者"] --> D["生存期仅数月至一年 / 致死率极高"] end subgraph CostBurden["每年 2500 亿美元抗癌开销"] E["NASA 全年预算的 10 倍"] F["教育部预算的 3 倍"] G["平均每个家庭分摊 2,000 美元"] end ``` --- ### [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) 破解沟通瓶颈：Dognosis 的多模态软硬件技术栈过去几十年，犬类癌症检测无法走出实验室的核心症结并非犬类的嗅觉极限，而是人犬沟通与规模化训练的瓶颈：传统模式往往需要耗费 12 个月以上训练驯犬员去理解犬类的微表情与动作反馈，不仅成本高昂且无法实现自动化与客观化。Dognosis 通过构建一套结合自动化硬件、计算机视觉、脑机接口与贝叶斯算法的多模态系统彻底移除了人类主观干扰。该系统的第一支柱是名为 **SniffSpace** 的自动化气味呈现与奖励平台。每个测试舱配置了红外与可见光摄像头，配合基于 Meta 开源架构 **V-JEPA 2** 微调出的自研计算机视觉模型 **Feral**，能够毫秒级判定犬只是否进行了一次高质量的有效嗅探，并精准锚定嗅探发生的时间切片。测试犬 Chloe 在进行检测时，头部佩戴名为 **DogSense** 的无创脑机接口（BCI）硬件以采集脑电（EEG）信号，身体则穿着采集体态语言、呼吸频率和心率变化的专用传感背心（Doggy Iron Man suit）。所有生物传感数据汇聚至名为 **DogOS** 的贝叶斯决策模型中，生成标准化的癌症风险评分。在全自动化流水线下，仅需一个由 30 只训练犬组成的核心犬群，每年即可完成 100 万份临床样本的检测。 ```mermaid flowchart LR accTitle: Dognosis 多模态检测技术架构 subgraph Sensing["第一层：多模态数据采集"] A["SniffSpace 视觉监控<br/>(红外/可见光)"] B["DogSense 脑机接口<br/>(非侵入式 EEG)"] C["犬用传感背心<br/>(呼吸/心率/体态)"] end subgraph Processing["第二层：AI 时空对齐与解析"] D["Feral 视觉模型<br/>(基于 V-JEPA 2 / 判定嗅探质量)"] E["多模态信号时间切片对齐"] A --> D D --> E B --> E C --> E end subgraph Decision["第三层：诊断输出"] F["DogOS 贝叶斯决策系统"] G["客观癌症危险度评分"] E --> F F --> G end ``` - **SniffSpace**：自动化样本分发与正向激励仓，完全实现无人化闭环。 - **Feral**：利用 Meta V-JEPA 2 视觉模型微调，精准评估嗅探动作的有效性与起止时间点。 - **DogSense & 传感背心**：无创脑电图（EEG）硬件结合呼吸、心率及动力学传感。 - **DogOS**：贝叶斯概率融合算法，将犬脑信号与外周生理指标转化为量化诊断结果。 - **通量指标**：30 只犬组成的作业群年检测通量可达 100 万份样本。 --- ### [11:29](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=689s) 临床试验验证与 2028 年家庭化早筛愿景现阶段癌症筛查依从性低的主要原因在于传统筛查过程繁琐痛苦（"feels like a chore"），患者往往需要专门请假、预约诊所排队并经历侵入性检查。Dognosis 提出的全新路径是将筛查转化为轻松的居家体验：用户收到寄送至家门口的采样盒，只需在早晨阅览报纸或玩 Wordle 填字游戏的 10 分钟内佩戴专用呼吸面罩，随后封存寄回中央实验室。若检测结果显示高反应性（High Reactivity Score），则直接进入针对性影像学确诊流程，该流程可完全对接医保报销。这一模式已通过全球规模最大的犬类嗅觉临床试验证实：Dognosis 联合印度知名医疗机构（包括 Narayana Health 与 ClearMedi Healthcare）完成了近 3300 例受试者的呼吸气味临床研究。该研究覆盖 7 大器官系统的 20 种癌症类型，结果表明犬类能识别全部测试癌种，整体特异性达到 **91.8%**，对 I 期癌症的敏感性是现有液体活检/机械检测手段的近 5 倍，而成本仅为其十分之一。 - **临床试验规模**：印度 3300 人临床验证，涵盖 7 大主要系统、20 个癌种（包括 I 至 IV 期）。 - **核心性能数据**：特异性 91.8%，早期敏感性约为传统检测仪器的 5 倍，成本降至约十分之一。 - **扩展路线图**： - 次年年中公布 30,000 人大规模随机对照试验结果（15,000 人常规标准护理 vs 15,000 人犬类呼吸分析）； - 设立美国得克萨斯州达拉斯总部； - 2027 年在美国开展本土临床验证； - 2028 年正式面向美国市场推出送货上门的无创居家早筛试剂盒。 ### [14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) 嗅觉超能力拓展：从气味视觉化到教 AI 学会“闻味” 在展示了癌症早筛的基础能力后，演讲者揭晓了另外两项更具颠覆性的“犬类超能力”。第二项超能力是将“气味地图”转化为人类可见的“视觉地图”。Dognosis 正在与新加坡警队合作开展试点项目，成功证明了系统可以将搜救犬和警犬在现场嗅闻到的气味实时转化为警员可视化的空间图像。演讲者用导航系统作了生动的类比：人类拥有 GPS，在拉斯维加斯可以精准定位并导航到最近的 In-N-Out 汉堡店，但人类在气味的世界中却完全是一片盲区。“这是对人类彻底封闭的一个维度，因为我们早早就认定嗅觉不如视觉重要；而借助这些工作犬，我们正为人类的认知打开全新的感官维度。” 第三项超能力则是借鉴了 AI 领域著名的基于人类反馈的强化学习（RLHF）机制。RLHF 曾是推动大语言模型实现质的飞跃、让 ChatGPT 真正具备实用价值的核心技术——通过人类持续给出的正负反馈来微调模型。Dognosis 正在构建完全相同的学习范式，只不过提供实时反馈的导师变成了犬类的大脑信号。通过犬类对微量气味分子的神经反应作为真值标注，团队正在建立一套全新的神经 AI 架构，致力于“教会 AI 如何去闻”。 ```mermaid flowchart LR accTitle: 基于犬类神经反馈的嗅觉AI训练流程 subgraph S1["生物感知层"] A["犬类嗅觉捕捉微量分子"] --> B["脑电信号 (EEG)"] end subgraph S2["解码与标注"] B --> C["犬脑状态实时解析"] C --> D["犬类反馈强化信号 (RLDF)"] end subgraph S3["模型进化层"] D --> E["嗅觉神经AI大模型"] E --> F["气味数字化 / 空间视觉地图"] end ``` - **新加坡警队试点**：将现场警犬嗅到的微量气味信息实时解码并投射到警用战术界面上。 - **嗅觉版强化学习**：利用犬类嗅球与大脑的直接电生理反应作为标签，训练 AI 具备分辨百万级挥发性有机化合物（VOC）的能力。 --- ### [16:34](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=994s) 癌症意识的历史启示：粉红丝带与早筛革命为了说明技术与社会认知结合的重要性，演讲者将视角拉回到八十多年前的乳腺癌倡导运动。今天普及的乳腺癌筛查并非一蹴而就，其源头可以追溯到 1940 年，当时麦克辛·戴维斯（Maxine Davis）在《好管家》（Good Housekeeping）杂志上发表了极具勇气的文章《让光照进癌症》（Let There Be Light on Cancer），公开打破禁忌，呼吁社会正视侵蚀女性健康的乳腺癌问题。接力棒随后在关键的历史节点上传递：1974 年，美国第一夫人贝蒂·福特（Betty Ford）打破白宫沉默，公开宣布自己确诊乳腺癌并成功抗癌，这在当时是前所未闻的壮举；1982 年，为了纪念因病逝世的妹妹，南希·布林克尔创立了苏珊·G·科曼乳腺癌基金会（Susan G. Komen Foundation）；直到戴维斯发文 52 年后的 1992 年，《Self》杂志才正式推出了如今闻名全球的“粉红丝带”标志。正是这场跨越半个世纪的公众意识革命，推动了乳腺 X 光钼靶筛查的广泛普及。自 1989 年死亡率达到峰值以来，乳腺癌患者的死亡率已大幅下降 50%，挽救了超过 50 万人的生命。演讲者强调，攻克其他类型的癌症同样需要这种强度的公众意识与行动主义。 - **1940 年**：麦克辛·戴维斯发表《Let There Be Light on Cancer》，开创癌症公开讨论先河。 - **1974 年**：第一夫人贝蒂·福特公开披露病情，彻底改变癌症病患的社会耻感。 - **1982 年与 1992 年**：苏珊·科曼基金会成立；《Self》杂志设计粉红丝带，奠定现代筛查意识基础。 - **临床成效**：自 1989 年峰值以来，乳腺癌死亡率下降 50%，逾 50 万人因早期筛查得以存活。 --- ### [18:51](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1131s) 现场演示：首只美国团队工作犬 Snow 与“萨拉米香肠刺激测试” Dognosis 目前在印度拥有一支由 50 名专业人员和 15 只训练有素的工作犬组成的科研与筛查团队。在演讲现场，演讲者隆重请出了美国本土工作犬团队的第一位成员——他自己的爱犬 Snow。当 Snow 欢快地走上讲台时，现场气氛达到高潮。演讲者笑着解释，为了不让 Snow 在台下众多观众面前感到紧张，他没有让它当场佩戴脑机接口头盔，而是展示了当天早晨刚刚采集的实测数据。屏幕上展示了清晰的脑电波波形图，被称为“萨拉米香肠刺激测试”。每当 Snow 闻到一块萨拉米香肠时，其 EEG 信号上就会出现一个极其鲜明而剧烈的特征峰值。演讲者指出，虽然现场没有携带癌症生物样本，但底层的神经生物学原理完全一致：“只要你能训练一只狗对某种气味感到兴奋，那种兴奋感就会直接穿透它的身体，体现在它的大脑神经元放电中。结合现代 AI 与脑电采集技术，我们终于能够精确读懂这种信号。”他动情地表示，终结癌症的希望并不仅仅寄托于 Reddit 上讨论的下一项颠覆性疗法或下一代化疗药物，而在于将普惠性早期筛查的意识推向大众。 - **团队规模**：印度基地常驻 50 名跨学科研究人员及 15 只经过严格医疗嗅探训练的工作犬。 - **萨拉米香肠测试原理**：气味刺激引发特异性嗅觉中枢神经元同步放电，在 EEG 信号中形成可被机器学习模型识别的高信噪比特征峰。 - **核心逻辑**：将犬类数亿年进化形成的极致嗅觉敏感度，通过无创脑电采集转化为客观、可量化的数字生物信号。 --- ### [20:55](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1255s) 创始团队与合作致谢：用集体力量推动癌症早筛落地在演讲的尾声，演讲者呼吁在场的观众拿出手机扫描屏幕上的二维码加入支持者名单。他真诚地坦言，公司已经顺利完成了种子轮融资，产品与算法也得到了临床验证，当下发起呼吁并不是为了募集资金，而是为了向医疗体系和监管机构证明公众对无创、可负担早筛技术的迫切需求。“决定这项技术最终能否走向世界的关键，在于我们人类自己是否愿意站出来表明立场。AI 的意义绝不仅限于开发下一个 App 或是训练下一个大语言模型，它应该用来做真正能改变生命的大事。” 演讲者回顾了公司的初创历程，特别致谢了两位联合创始人——来自加州大学伯克利分校的认知科学家 Ákos 以及退役军犬专家 Itamar，正是他们当初看似“疯狂”的设想奠定了 Dognosis 的基石。此外，他还向敢于开展前沿临床试验的多家医疗健康中心、提供关键脑电硬件支持的 g.tec 脑机接口团队，以及现场正在为这项技术拍摄纪录片的 Harvey Nash Tech 的 David Savage 表达了诚挚的感谢。 - **联合创始人背景**：加州大学伯克利分校认知科学家 Ákos 与前军犬特训专家 Itamar。 - **关键技术合作伙伴**：奥地利脑机接口研发机构 g.tec（提供核心 EEG 采集与传感器工程支持）。 - **行动呼吁**：收集用户意向名单以展现真实市场与患者需求，从而打破医疗早筛准入的传统壁垒。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=4s) 演化史上最关键的奇迹发生在 3 万年前的欧亚大陆——人类首次与狼相遇。 — Amir Alpert > [02:20](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=140s) 犬类具备感知异常的能力，但缺乏精准、客观的沟通渠道向人类传达预警。 — Amir Alpert > [07:15](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=435s) 过去几十年，犬类癌症检测无法走出实验室的核心症结并非犬类的嗅觉极限，而是人犬沟通与规模化训练的瓶颈。 — Amir Alpert > [11:29](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=689s) 现阶段癌症筛查依从性低的主要原因在于传统筛查过程繁琐痛苦（feels like a chore）。 — Amir Alpert > [14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) 这是对人类彻底封闭的一个维度，因为我们早早就认定嗅觉不如视觉重要；而借助这些工作犬，我们正为人类的认知打开全新的感官维度。 — Amir Alpert > [14:58](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=898s) 通过犬类对微量气味分子的神经反应作为真值标注，团队正在建立一套全新的神经 AI 架构，致力于教会 AI 如何去闻。 — Amir Alpert > [16:34](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=994s) 自 1989 年死亡率达到峰值以来，乳腺癌患者的死亡率已大幅下降 50%，挽救了超过 50 万人的生命。攻克其他类型的癌症同样需要这种强度的公众意识与行动主义。 — Amir Alpert > [18:51](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1131s) 只要你能训练一只狗对某种气味感到兴奋，那种兴奋感就会直接穿透它的身体，体现在它的大脑神经元放电中。结合现代 AI 与脑电采集技术，我们终于能够精确读懂这种信号。 — Amir Alpert > [18:51](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1131s) 终结癌症的希望并不仅仅寄托于 Reddit 上讨论的下一项颠覆性疗法或下一代化疗药物，而在于将普惠性早期筛查的意识推向大众。 — Amir Alpert > [20:55](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1255s) AI 的意义绝不仅限于开发下一个 App 或是训练下一个大语言模型，它应该用来做真正能改变生命的大事。 — Amir Alpert > [20:55](https://www.youtube.com/watch?v=4FoUiWt8Cm8&t=1255s) 决定这项技术最终能否走向世界的关键，在于我们人类自己是否愿意站出来表明立场。 — Amir Alpert ## Mentioned Amir Alpert, Dognosis, Medical Detection Dogs, Daisy, 《我在雨中等你》（The Art of Racing in the Rain）, Enzo, NASA, SniffSpace, Meta V-JEPA 2, Feral, DogSense, DogOS, Chloe, Snow, Narayana Health, ClearMedi Healthcare, 新加坡警队, Maxine Davis, 《好管家》（Good Housekeeping）, Betty Ford, Susan G. Komen Foundation, 《Self》杂志, Ákos, Itamar, g.tec, David Savage, Harvey Nash Tech ## Video description Amir Alpert - Head of US Business Development, Dognosis What if the most sensitive Stage I cancer test on Earth isn't a lab assay — it's a dog's brain, decoded by AI? Dogs have been smelling cancer for decades; we taught AI to read what they smell. At Dognosis, we've built a clinically validated cancer screen by combining trained detection dogs, brain-computer interfaces, and a neural network that translates canine olfaction into digital diagnoses - 90.6% Stage I sensitivity, 91.3% specificity, peer-reviewed across 3,000 patient samples and roughly 5x more sensitive than liquid biopsy. The talk will walk through the technical stack (BCI hardware → EEG signal → neuroAI translation), the clinical pathway from canine signal to confirmed diagnosis, and the business model for at-home cancer screening at a price point that doesn't bankrupt patients. If you build AI for healthcare, work on multi-modal sensing, or want to see neuroAI deployed in the real world, this is the session. [Source](https://www.youtube.com/watch?v=4FoUiWt8Cm8)

Source quotes

[0:04] Amir Alpert: 演化史上最关键的奇迹发生在 3 万年前的欧亚大陆——人类首次与狼相遇。 [2:20] Amir Alpert: 犬类具备感知异常的能力，但缺乏精准、客观的沟通渠道向人类传达预警。 [7:15] Amir Alpert: 过去几十年，犬类癌症检测无法走出实验室的核心症结并非犬类的嗅觉极限，而是人犬沟通与规模化训练的瓶颈。 [11:29] Amir Alpert: 现阶段癌症筛查依从性低的主要原因在于传统筛查过程繁琐痛苦（feels like a chore）。 [14:58] Amir Alpert: 这是对人类彻底封闭的一个维度，因为我们早早就认定嗅觉不如视觉重要；而借助这些工作犬，我们正为人类的认知打开全新的感官维度。 [14:58] Amir Alpert: 通过犬类对微量气味分子的神经反应作为真值标注，团队正在建立一套全新的神经 AI 架构，致力于教会 AI 如何去闻。 [16:34] Amir Alpert: 自 1989 年死亡率达到峰值以来，乳腺癌患者的死亡率已大幅下降 50%，挽救了超过 50 万人的生命。攻克其他类型的癌症同样需要这种强度的公众意识与行动主义。 [18:51] Amir Alpert: 只要你能训练一只狗对某种气味感到兴奋，那种兴奋感就会直接穿透它的身体，体现在它的大脑神经元放电中。结合现代 AI 与脑电采集技术，我们终于能够精确读懂这种信号。 [18:51] Amir Alpert: 终结癌症的希望并不仅仅寄托于 Reddit 上讨论的下一项颠覆性疗法或下一代化疗药物，而在于将普惠性早期筛查的意识推向大众。 [20:55] Amir Alpert: AI 的意义绝不仅限于开发下一个 App 或是训练下一个大语言模型，它应该用来做真正能改变生命的大事。 [20:55] Amir Alpert: 决定这项技术最终能否走向世界的关键，在于我们人类自己是否愿意站出来表明立场。

Source quotes

Amir Alpert: What if the most sensitive Stage I cancer test on Earth isn't a lab assay — it's a dog's brain, decoded by AI? Amir Alpert: Dogs have been smelling cancer for decades; we taught AI to read what they smell. Amir Alpert: At Dognosis, we've built a clinically validated cancer screen by combining trained detection dogs, brain-computer interfaces, and a neural network that translates canine olfaction into digital diagnoses [7:15] Amir Alpert: The fundamental barrier to utilizing canine detection in clinical diagnostics has never been the dogs' olfactory acuity; it has been the communication bottleneck and lack of scalability. [7:15] Amir Alpert: doggy Iron Man suit [16:34] Maxine Davis: Let There Be Light on Cancer [18:51] Amir Alpert: salami stimulus test [20:14] Amir Alpert: The path to ending cancer is not about the next therapy. It is not about the next breakthrough that we see on Reddit. It is not about the next article, about the next chemotherapy or the next amazing medical improvement. It is about awareness.

Original transcript

[0:04] I'm here representing Dognosis. And at Dognosis, we believe canines have [0:10] superpowers. And we believe our role is to find a way to unleash these superpowers for [0:16] the betterment of humanity. 300,000 years ago, when man first evolved from primates, [0:24] they moved around. We were a very curious species. 60,000 years ago we moved into India. 50,000 years [0:34] ago we moved into Europe. 20,000 years ago we crossed over from Russia into the Americas. But [0:44] what I want to talk about today is what happened 30,000 years ago in modern day Eurasia. I don't [0:52] think it's a stretch of the imagination to say what happened here was a miracle. [1:00] What happened 30,000 years ago is that the human first met the wolf. This [1:06] was a special relationship. As humans easily recognize the wolf's potential, [1:13] we used them to help us hunt animals that were bigger than us. As we developed our own camps, [1:21] the wolves protected us. And there's even a well-known theory that it was the wolf's [1:28] ability to guard us at night that enabled all of us to start to sleep together at the same time [1:34] and to start to dream. Without a doubt though, the most important thing that happened in this [1:41] relationship is that the wolf and the human became friends. We developed a bond, a companionship, and [1:50] likely half of you in this audience continues to share with this bond in the dogs you have at home [1:59] today. They continue to serve the same roles. We still use these canines for hunting. We still have [2:07] canines protecting us. And without a doubt, we still have canines doing even the most bizarre [2:15] things as our friends. The reality is that these dogs can do so much more. As somebody develops [2:24] cancer, the metabolism in their entire body shifts. Their immune system begins to fight [2:30] this disease. The changing in the metabolism leads to molecules releasing into our bloodstream. The [2:39] blood goes to our lungs and these molecules begin to come out in our breath. For 40 years, we've [2:47] known that dogs can detect cancer by smelling these volatile organic compounds. And I have to [2:57] give a special shout out to medical detection dogs out of the UK. This is a pioneering group [3:03] that in 2004, the founder of this group had a dog named Daisy and that dog diagnosed the founder [3:12] with breast cancer. That founder then decided that this special ability had to be shared. And [3:20] for the last 21 years, they have pioneered the foundation for what we're building off of today. [3:26] This phenomenon has been known for so long that we've even started creating books and movies [3:31] about it. In Racing in the Rain, it's the story of a narrating dog named Enzo who starts to feel [3:40] that something is wrong with the owner's wife. The wife had breast cancer. Unfortunately, Enzo never [3:49] had an ability to tell the owner or the wife what was wrong, and she passed away. It's a story about [3:59] a dog that knew that something was wrong, but did not have the ability to communicate it to the ones [4:04] that needed its help. I told you that we evolved 300,000 years ago. That's when the first homo [4:09] sapien emerged. But the reality is that cancer has been in us before homo sapiens even existed. [4:18] 1.7 million years ago. That's the first fossilized evidence of a tumor osteocaroma in a toe fossil in [4:26] South Africa. Think about that for a second. That means this disease that everyone in this [4:34] audience without a doubt has either experienced themselves unfortunately or know somebody that [4:40] has experienced it is a disease that has been with humans before humans are even humans. [4:49] [snorts] There's three numbers I want you to remember from this slide. The first number is [4:56] 2 million. Every year in the US, 2 million people will be diagnosed with cancer. 1 million of them [5:06] will be diagnosed with an early stage cancer where the modern therapies can actually make [5:12] some meaningful impact, give them a second chance. The other 1 million, their diagnosis [5:21] will be essentially a death sentence. Modern therapies will maybe extend it for a few months, [5:27] maybe a year, but ultimately they know that they will succumb and they will pass away. [5:36] The state of the house, as we want to call it, is that we spend [5:40] $250 billion every year to fight this disease. For some of you in the audience raising funds, [5:49] maybe this number is even understandable. For us at Dognosis, this number is so big it's difficult [5:54] to even comprehend. So, let me translate that. $250 billion is 10 times what we spend at NASA. [6:04] It's three times the amount of money we spend in the entire Department of Education. It's [6:10] $2,000 for every household in America. And the reality is we're not using that money very wisely. [6:22] 300,000 years ago, humans evolved. 30,000 years ago, the era of the working wolf started. 1,000 [6:33] years ago, the era of the friendly dog, when we started taking wolves and making new breeds. For [6:42] 40 years, we've known that canines can detect cancer. And it is only in the last 5 years with [6:50] the emergence of AI, with the emergence of the tools that so many of you know so well, with the [6:55] increased compute that we can start to bring this amazing ability, this legitimate superpower to the [7:02] world. That's why superpower number one atnosis is multic-cancer early detection for humans. [7:15] The problem has never been the dogs. The dogs have always been able to detect when something [7:22] is wrong. Fundamentally, this has been an issue in communication. It has been an issue in scaling [7:29] their ability because it has traditionally taken 12 plus months to train a human to understand [7:36] what the dog is thinking. It has taken months to train the dog to do its role and fundamentally [7:47] this has led society to say this just cannot scale. This ability, this superpower is not [7:54] something we can bring to the world. So it has remained in the academics. It has reigned, [7:58] it has remained as anecdotal evidence. But I'm here to tell you that we have solved this problem. [8:07] Half of the solution to this is what we call sniffspace. This is an automated [8:14] sample presentation and reward platform that takes the human out of the loop. Every bay has [8:21] infrared cameras, visual cameras that enables us to determine when a dog is accurately smelling a [8:27] sample and to understand objectively whether or not they think that sample is cancerous. [8:39] This is our dog Whiskey. [8:44] At dognosis, we built Feral. Feral was based on the pre-trained model by Meta, [8:51] an open- source model called VJEPA 2. What we did on top of that is we took this basic model, [8:59] and we ramped it up with the ability to determine when a sniff was a quality sniff. [9:06] A very obscure use of AI nonetheless, but a very important piece. With that, we're able to use [9:14] computer vision to accurately determine when a sniff is of sufficient quality to determine the [9:21] time and space when a dog can accurately determine whether a sample is cancerous. [9:30] Let me introduce you to Chloe. Chloe is smelling a breath sample from somebody [9:37] that breathed into a face mask for 10 minutes. What's clear though is Khloe is not doing this [9:43] alone. On top of Khloe's head, she's wearing a non-invasive brain computer interface that [9:49] we call dog sense. This is our hardware layer. On Khloe's body, she's wearing a doggy Iron Man suit. [9:57] This is measuring her body language, her respiration, her heart rate. These are all [10:04] the signals that help us to understand whether Khloe thinks that this sample is dangerous. [10:12] We have a software layer called dog OS. This is a Beijian model that enables us to integrate all [10:17] of these signals in an objective format and give a reliable determination about the danger of the [10:25] sample. taking the human out of the loop. Combining these three pillars of doggosis, [10:35] we are now able to scale this superpower so that a pack of 30 dogs can test 1 million samples a year. [10:46] What does that look like in real life? [10:51] At the top of this, we have feral. This enables us to determine the moment in [10:57] time that we should be looking at the signal. This enables us to take a time [11:02] slice and match it up to the EEG data looking into the dog's brains. Match it [11:08] up to the respiration and the movement data from the doggy Iron Man suit. Taken together, [11:17] this creates a multimodal approach to canine communication for early cancer detection. [11:29] Why do we not get cancer screens? We all know that we should do it. We all know that it [11:35] could save our lives. We all know people that have passed away that it could have [11:38] saved. And the reality is we don't get cancer screens because they feel like a chore. It's [11:45] thinking that we have to take time away from our lives, from our family, from our friends, [11:50] from our work. We have to make an appointment to go to a doctor's office that we do not want to go [11:54] to to get a result that we do not want to hear. So, what if it could be easier? The vision that [12:02] we're presenting here today is multicancer early detection that does not feel like a chore. You [12:08] wake up, there's a kit at your doorstep. The kit contains a simple breath mask like we've [12:15] gotten so used to from co. You breathe into this breath mask while reading the New York Times, [12:20] doing your whle on the couch for 10 minutes. You put that breath mask back into the kit. You mail [12:26] it back to us. And at a centralized lab, a pack of dogs that is well trained with the [12:30] right technology is able to determine whether or not you're in danger. If you're in danger, [12:36] if you have this high dog reactivity score, you then move into a workflow of imaging diagnostics [12:43] so that your physician has an actionable result that your insurance has to cover. [12:54] This isn't theoretical. At Dnosis, [12:59] we've completed the world's largest K9Action cancer clinical trial. [13:06] In this trial in India, we took nearly 3,300 participants in collaboration with [13:12] the health institutes in India, Neurania, Clear Media Healthcare. In this trial, [13:20] individuals just breathed into a mask for 10 minutes. This trial covered 20 cancer [13:27] types across seven major cancer groups. And I'm here to tell you, [13:30] we have not found a single cancer that these dogs cannot detect. [13:37] Here's the results. Across stage 1, stage two, stage three, stage four. [13:43] These dogs can tell you when you're in danger. With an overall specificity of [13:49] 91.8%. You do not have to be an expert, but I'm telling you, this is five times better [13:54] than any machine that has ever been created for cancer detection. at about a tenth of the cost. [14:06] Where are we now and where are we going? The first trial was 3,300 participants. By [14:12] the middle of next year, we'll be releasing results from a trial nearly 10 times bigger, [14:18] a randomized trial comparing 15,000 people with standard of care and 15,000 people that got the [14:24] breath analysis done by our canines. But this is not a technology that it will remain in India. We [14:32] have already started the process of bringing it to the US. We're launching our diagnosis [14:37] headquarters in Dallas, Texas. In 2027, we will start the clinical validation showing [14:42] that the same results that worked in India will work here in the US as well. And by 2028, [14:48] you will be able to order a kit to your doorstep to determine whether you're in danger. [14:58] That was superpower number one. And if you can believe it, superpower number [15:02] two and superpower number three are even bigger ideas. Superpower number two is [15:08] converting smell maps into visual maps. This is conducted with a pilot with the [15:15] Singaporean police units where we've been able to demonstrate that we can take what [15:21] a dog smells in the field and show it to the police to see what it is that's out there. [15:29] I'm going to leave you with this one idea around this. We have a global positioning system, [15:33] right? We have GPS tells us that we're in Las Vegas right now. Tells us how to get from here to [15:38] the next In-N-Out Burger. We have no idea where we sit in the smell of the world. This is a dimension [15:46] that is essentially locked to humanity because we have decided that smell is less important than [15:50] vision. And with these dogs, we can bring that new dimension to our understanding. [16:01] Superpower number three. Probably everyone in this audience knows about RHF, right? [16:08] This is essentially the process that unlocked chatbt that took it from chatbt 2 to chatbt [16:13] 3. This was human feedback telling the AI what was good and what's wrong. We're doing [16:19] the exact same thing at doggnosis except the dogs are telling us what's right and [16:23] what's wrong. With this framework, we think we can teach AI to smell. [16:34] As I wrap up this call, I want to go back a long time. [16:41] The movement in breast cancer as we know it today did not start recently. We can trace its origins [16:49] to 1940 when Maxine Davis in Good Housekeeping magazine at the time took a very courageous step [16:57] to say we need to talk about cancer. It is affecting the women in our society. Why are [17:02] we not talking about breast cancer? She wrote an article called Let There Be Light on Cancer. [17:11] Following in her footsteps, it was 1974 when Betty Ford, the first lady at the time, came out [17:17] publicly for the first time and said, "I have been diagnosed with breast cancer and I have survived." [17:23] Today, that seems like something we would automatically do. In 1974, it was unheard of. [17:31] In 1982, it was the death of a sister, Susan G. Kman, that inspired her surviving sister to [17:40] start the Susan G. Kman Foundation, one of the largest breast cancer foundations today. And it [17:47] was not until 1992, 52 years after Maxine Davis, that we got the pink ribbon that we know today. [17:56] This pink ribbon was created by Self magazine, another woman's magazine, another part of the [18:01] women's movement. And it's this pink ribbon today that drives breast cancer awareness month, drives [18:11] the women to get mammogs at a exceedingly high rate compared to other cancer screens. And it's [18:18] the same type of activism that will be required to tackle the rest of the cancers. This movement in [18:26] breast cancer accomplished amazing results. Since its peak in 1989, we have seen a 50% decrease in [18:37] mortality rate if you get diagnosed with breast cancer. That's over half a million lives saved. [18:51] All of this is done in India with an amazing team. 50 humans, [18:56] a pack of 15 dogs, now unleashing K9 superpowers. [19:06] And today I want to introduce you to the first dog [19:09] in our USA pack. He's a dog that means a lot to me as he's my own dog. Snow, [19:17] you want to come on stage? Yeah. Yeah. Come on. Let me introduce you to Snow. [19:30] I didn't make him wear the helmet because maybe he'd be a little nervous with all [19:33] the faces. But I will tell you that this morning we did the salami stimulus test. [19:43] Every one of those peaks that you see is Snow's EEG as he smells a piece of salami. I didn't have [19:52] enough cancer samples to bring with us. It's the same fundamental principle though. If you [20:00] can train a dog to be excited about something, that excitement comes through in their body. [20:05] It comes through in their brains. And with AI now and the right technology, we can understand that. [20:14] Good job, boy. The path to ending cancer is not about the next therapy. It is not [20:19] about the next breakthrough that we see on Reddit. It is not about the next article, [20:22] about the next chemotherapy or the next amazing medical improvement. It is about awareness. [20:30] Like I said, every one of you in this audience, me included, [20:32] knows somebody that was diagnosed. And unfortunately, some of you in [20:36] this audience may have been diagnosed. I'm asking you today to take out your phones. [20:55] Just scan this QR code. [20:58] [snorts] We don't need money. We've raised our seed. We know we have a good product. [21:07] The fundamental thing that will make this either enter the world or not is [21:11] whether us as humans take a stand that this is needed. It's time that we tell [21:15] the world some of the great things that can come from AI. It's not about the next app. It [21:21] doesn't have to be the next LLM model. It can be something that makes such an [21:24] impact. And I'm just really thrilled to be able to show that to you guys today. [21:33] With that, I just want to give an amazing appreciation for all of the organizations [21:37] that have helped us along the way. This idea started with two founders, [21:44] cognitive scientists out of Berkeley named Akos and a former K-9 military specialist named Edomar. [21:54] At the time when they started this, it was a crazy idea. [22:01] In addition, we have all of the healthc care centers that have helped us that took a chance [22:05] to start a clinical trial that seemed out of this world. We have our collaborators [22:12] at GTech that helped us with the brain computer interface, the EEG technology that is essential [22:19] to understanding what's going on in these dogs heads. And I want to give a big shout [22:25] out as well on David Savage in the back there from Harvey Nash Tech. who took the chance to [22:32] reach out to us and say that they wanted to do a documentary on what we're doing here. [22:39] I'm going to leave this up because I know not everybody here took out their phones [22:43] and that's okay. I know it's weird. If you take out your phones, you scan this QR code, [22:49] we will only ask for your name, some brief demographic information so we can put you on [22:53] a list. So the next talk that we give, we can say our list is growing. We can [22:58] say people are taking a stand. They want this product. And with your awareness, [23:04] with your interest, we unlock doors that are can seem insurmountable.