👨‍🎓Meet the AI Student Who Never Sleeps

Hey TEDchats Fam!

Happy Thursday! Once upon a thesis crisis, every biology grad student has whispered to themselves, “If only I had a clone to read all these papers”. Well, FutureHouse just made that clone—only it runs on GPUs and doesn’t need coffee breaks.

🧑‍💻 WEEKLY PROMPT

🍕 QUICK BITES

Meet Finch — The AI Grad Student for Biology

FutureHouse, a nonprofit bringing superintelligent AI to science, has launched a new tool called Finch. The AI mimics a first-year biology grad student by analyzing research papers and running code to produce scientific insights. It’s designed to support “data-driven discovery in biology” and has already been used for tasks like RNAseq analysis. While promising, Finch is still in closed beta and prone to errors—prompting FutureHouse to bring in human experts to refine it. Backed by Eric Schmidt, FutureHouse has also just launched a broader platform to bring the first publicly available superintelligent scientific agents to researchers via web and API (see FutureHouse’s announcement).

OpenAI Restructures Amid Legal Pressure and Investor Demands

OpenAI is revamping its corporate structure, converting its for-profit arm into a public benefit corporation (PBC) still controlled by its nonprofit board. The move follows pressure from regulators and stakeholders, including Microsoft and attorneys general in California and Delaware. While the change could open the door to an Initial Public Offering (IPO), experts warn the nonprofit’s grip on key technology may limit investor influence. OpenAI maintains it has no immediate plans to go public, though the new PBC structure would allow for it in the future (see OpenAI’s announcement).

Reddit Cracks Down After AI Bots Fool Thousands!

A wave of outrage hit Reddit last week after news broke that researchers had conducted a covert experiment using AI bots disguised as real users. Without disclosure, the bots infiltrated the “Change My View” subreddit, posting over 1,700 comments while posing as individuals with emotionally and politically charged identities. In response, CEO Steve Huffman announced plans to partner with third-party verification services to confirm users are real people — a move that could upend Reddit’s long-standing culture of anonymity, though the company insists it won’t collect names or personal identifiers (read apology statement).

🧰 AI TOOLBOX 

🤯 MIND SPARK

💥 Gemini 2.5 Pro Gets a Powerful I/O Upgrade: Google has launched Gemini 2.5 Pro Preview (I/O edition), a major upgrade with top-ranked coding and web dev skills. Now #1 on WebDev Arena, it replaces the previous model at no extra cost via Gemini API, Vertex AI, AI Studio, and the Gemini app — just ahead of the I/O conference on May 20–21 (see Google’s announcement).

🇺🇸 Top Tech Leaders Urge AI Education: Over 250 tech leaders, including Microsoft CEO Satya Nadella, have signed an open letter urging U.S. schools to make AI and computer science core subjects. The letter warns that the U.S. is falling behind countries like China and South Korea, and emphasizes the need to train students to create AI — not just use it (see the open letter).

🧬 Anthropic Launches AI for Science Program: Anthropic has launched an ‘AI for Science’ program to support research in biology and the life sciences, offering up to $20K in Claude API credits (see rules). The program prioritizes projects that tackle complex biological systems, analyze genetic data, or accelerate drug discovery, with winning applications selected monthly based on scientific merit and biosecurity considerations (see Anthropic’s announcement).

🏛️ AI R&D Faces Deep Cuts as IBM Pushes Back: IBM CEO Arvind Krishna, in a TechCrunch interview, warned against steep federal cuts to AI R&D, including a proposed 50% slash to NSF funding and threats to the CHIPS Act. IBM has reportedly lost $100M in federal contracts, but Krishna hopes investment will rebound within a year (read more).

🩺 Chatbots and Health Checkups: A Risky Combo? An Oxford-led study found that AI chatbots like ChatGPT may hinder health decision-making. In tests with 1,300 users, participants often missed key details and misjudged condition severity, while chatbot responses mixed helpful and harmful advice. Experts say the tools aren’t ready for high-risk medical use (see the study).

🌍 France’s AI Darling Mistral Expands Into the Enterprise Arena: French startup Mistral AI captured international attention in February when its chatbot Le Chat reached 1 million downloads in just two weeks. Now, it’s expanding into the enterprise space. Le Chat Enterprise, powered by the new Mistral Medium 3 model, delivers a unified AI platform for organizational work (see Mistral’s announcement).

🎨 AI ART