Genomic Foundation Model
Highlights
Insights
1. AI systems have been developed to mimic expert judgment in various domains.
2. Businesses may consider three waves of AI adoption: enhancing efficiency, elevating outcomes, and creating transformative systems.
3. AI tools can be used for long-term support for our work (like exoskeletons), or to enhance independence and learning (like scaffoldings).
News
1. Microsoft Ignite 2024 introduced powerful AI agents and enhancements for Microsoft 365 Copilot.
2. ElevenLabs now offers ability to build customizable conversational AI voice agents.
3. Pixtral Large is a 124B open-weights multimodal model that demonstrates frontier-level image and text understanding.
4. AI can now create highly accurate replicas of individual personalities from just a two-hour interview.
5. A study found that ChatGPT outperformed doctors in diagnosing illnesses from medical case histories.
6. Evo, a 7-billion-parameter genomic foundation model, leverages AI to predict and generate genetic sequences from DNA to whole genomes.
Innovation Insights
1. How 3 companies are using AI to mimic expert judgement (Harvard Business Review)
Bavaria Ventures employed an AI system to evaluate startup applications, streamlining processes and improving accuracy by mimicking senior analysts’ assessments. Sat Vista Analytics developed AI “avatars” of their top specialists, automating satellite data analysis and freeing experts for complex tasks. Be Smarter implemented AI agents to upskill R&D employees, enabling tailored problem-solving assistance and reducing training costs. These cases highlight AI’s ability to transfer and enhance expertise, creating agile, responsive, and efficient organizational systems.
2. What ChatGPT’s corporate victims have in common (The Economist)
Businesses hit hardest by generative AI, like Chegg, Stack Overflow, and RWS, face challenges when AI can easily replicate their services with minimal customization. Industries with low risks from AI errors, such as education or translation, are particularly vulnerable, as the consequences of mistakes are minor compared to fields like legal or medical work. Companies in the AI “firing line” can mitigate risks by adopting innovative AI-driven offerings, as seen with Duolingo’s AI-powered language tools that have won investor praise. Ultimately, adaptability and creative integration of AI are key to surviving and thriving in the face of technological disruption.
3. Shaping your AI strategy – 4 foundational perspectives (Board of Innovation)
To shape an effective AI strategy, businesses must navigate the three waves of AI adoption—enhancing efficiency, elevating outcomes, and creating transformative systems—while planning for long-term innovation. The plummeting costs of digital production enable unprecedented opportunities, from real-time personalization to scalable AI agents, reshaping industries and consumer experiences. The convergence of artificial intelligence and biological advancements promises transformative impacts on health, memory, and human augmentation. To gain a competitive edge, businesses must strategically leverage AI to reinforce strengths, innovate unique offerings, or redefine business models, ensuring resilience and leadership in an AI-driven world.
4. Superhuman Knowledge Workers? AI Exoskeletons and Scaffoldings (Gianni Giacomelli)
This article explores the concept of AI-powered tools like exoskeletons and scaffoldings, which augment human capabilities and productivity in knowledge work. Scaffolding provides temporary support to foster independence and learning, while exoskeletons offer ongoing augmentation for cognitive and decision-making tasks. Research shows tools like MIT’s Supermind Ideator significantly enhance creativity and efficiency compared to general-purpose AI, though human judgment remains crucial for refining outcomes. By designing AI systems with a focus on workflows, collaboration, and domain-specific needs, these tools can unlock new levels of productivity, much like Excel transformed work decades ago, fostering innovation and adaptability across industries.
AI Innovations
1. OpenAI
OpenAI is bringing ChatGPT’s Advanced Voice Mode to the web, enabling users to interact with the AI chatbot through voice directly in their browser (TechCrunch).
OpenAI’s latest GPT-4o update significantly enhances ChatGPT’s creativity, making it more engaging, natural, and competitive (Tom’s Guide).
2. Meta
Messenger introduces AI-generated backgrounds, HD video calls, noise suppression, audio/video messaging, and hands-free functionality with Siri, enhancing the calling experience (Meta).
3. DeepSeek
DeepSeek’s new R1-Lite-Preview reasoning model demonstrates superior performance to OpenAI’s o1-preview, excelling in logical reasoning and transparency, with plans for open-source releases to follow (VentureBeat).
4. Microsoft
Microsoft Ignite 2024 introduced powerful AI agents and enhancements for Microsoft 365 Copilot, enabling automation, improved collaboration, and productivity tools tailored for hybrid work environments (ZDNet).
5. Alphabet
Google is introducing a feature for Gemini that allows it to “remember” specific details you’ve shared, like your interests and personal goals (The Verge).
6. Anthropic
Claude AI now has the capability to review your Google Docs and provide feedback (ZDNet).
7. Perplexity
Perplexity’s new AI-powered shopping assistant simplifies online shopping with features like one-click checkout for Pro users, Snap to Shop visual search, unbiased product recommendations, and integrations with platforms like Shopify, while also introducing a Merchant Program to enhance product visibility and streamline commerce experiences (Perplexity).
8. ElevenLabs
ElevenLabs has launched a platform for building customizable conversational AI agents, allowing users to integrate knowledge bases, fine-tune voice and response parameters, and choose language models for tailored interactions (TechCrunch).
9. Mixtral
Pixtral Large, a 124B-parameter multimodal model built on Mistral Large 2, excels in understanding complex visual and textual data, achieving state-of-the-art performance, while offering enterprise-ready features for tasks such as multilingual OCR, chart analysis, and advanced reasoning with a 128K context window (Mixtral).
10. H
H, the Paris AI startup, has launched Runner H, an agentic AI for process automation and quality assurance (TechCrunch).
11. Personality
AI can now create highly accurate replicas of individual personalities from just a two-hour interview, enabling applications from research simulations to personalized digital twins (MIT Technology Review).
12. Robotics
Advances in physical AI are enabling robots to perform complex real-world tasks, bringing the long-envisioned era of highly capable, adaptable, and interactive machines closer to reality, despite ongoing challenges in safety, adaptability, and scalability (Financial Times).
13. Model training
Chinese AI company 01.ai trained a high-performing GPT-4 rival, Yi-Lightning, using just $3 million and 2,000 GPUs by employing innovative engineering optimizations, highlighting cost-efficiency compared to OpenAI’s $80–$100 million investment (Tom’s Hardware).
14. Spotify
Spotify is embracing generative AI tools like Suno and NotebookLM to enhance user experiences by enabling creators to use AI for music and podcast production, while maintaining human creativity and compliance with copyright laws (Alex Kantrowitz).
15. Suno
Suno’s upgraded AI music platform, Suno v4, enhances song creation with improved vocals, dynamic lyrics, and versatile features like Personas and Covers, appealing to both casual users and professional musicians (TechRadar).
16. AI for science and medicine
The MIT Jameel Clinic has released Boltz-1, an open-source biomolecular interaction modeling tool achieving AlphaFold3-level accuracy, providing researchers worldwide with a powerful and accessible platform for advancing drug design and structural biology (MIT).
Stanford researchers developed SEQUOIA, an AI tool that predicts tumor gene activity from standard biopsy images, potentially replacing costly genomic tests and accelerating cancer treatment decisions (Stanford).
Evo, a 7-billion-parameter genomic foundation model, leverages AI to predict and generate biological sequences from DNA to whole genomes, enabling groundbreaking tasks like designing functional CRISPR systems and advancing biological engineering across molecular and genomic scales (Science).
A study found that ChatGPT outperformed doctors in diagnosing illnesses from medical case histories, revealing both the chatbot’s superior diagnostic accuracy and the challenges doctors face in effectively utilizing AI tools and overcoming biases in their own diagnoses (New York Times).
The NIH-developed AI algorithm, TrialGPT, streamlines matching patients to clinical trials by analyzing eligibility and summarizing criteria, saving clinicians time and accelerating research progress (NIH).
BiomedParse, an advanced foundation model for biomedical image analysis, unifies object recognition, detection, and segmentation using natural language prompts, achieving state-of-the-art performance across nine imaging modalities and enabling more efficient and accurate clinical insights (Microsoft).
17. Gaming
Genies launched “Parties,” a customizable AI-powered platform that enables brands and creators to build immersive, user-generated gaming ecosystems driven by smart avatars and personalized experiences (VentureBeat).
Other Innovations
1. Human cell atlas
The Human Cell Atlas, a groundbreaking global effort to map all human cell types, has already cataloged 62 million cells, offering vital insights into health and disease (Nature).
2. Malaria vaccine
Scientists have developed a malaria vaccine delivered through genetically modified mosquitoes, which boosted immunity and prevented infection in nearly 90% of trial participants (Nature).
