FDA plans to speed up scientific reviews using AI

Welcome to the prompt.

AI is finding its path to the public sector, with its latest entry point being the US Food and Drug Administration, which yesterday announced it had deployed an agency-wide generator AI tool called ELSA. The FDA will use AI systems to find high-priority testing targets, generate code and create a database, and summarise information from safety assessments and scientific reviews.

So let's get into the headline.

Big play

Meta, generation 97% of advertising revenue Currently developing on Facebook and Instagram ai Tools aimed at supporting brands Create and target your ads entirelyAccording to Wall Street Journal. I'm also planning on using AI Personalize your ads By tweaking them for different users based on factors such as location and budget.

Market movements

Chip Giant Nvidia reported $44.1 billion It broke Wall Street forecasts with revenue for the most recent quarter, ending April 27th. However, the company's annual revenues have declined. Forbes Mostly it has been reported A $50 billion hit after Trump administration halted exports of H20 AI Tip To China.

Data dilemma

Academic Journal And there is a scientific website I was overwhelmed by the bot According to the company, it carries hundreds of millions of documents and photos to train generative AI models. Nature. Active bot scraping activities have pushed several sites to the brink Close their doors, By surge in costs thanks to server overloading Slow site functionality for legitimate users.

Deep Diving

Snorkel AI CEO Alex Ratner remembers a time when data labeling (the tough task of adding context to the range of raw data and scoring the responses of AI models) was considered the “cleaning” task of AI researchers. But that quickly changed in 2022 when ChatGpt surprised the world, inhaling new lives (and billions of dollars) and infused a series of startups rushing to train capable models by providing human sign data to things like Openai and humans.

It appears that the busy field of data labeling is currently experiencing another shift. Few companies train large language models from scratch, leaving the tasks behind at high-tech giants instead. Instead, they are fine-tuning models, building applications, and creating demand for specialized data in areas such as software development, healthcare, and finance. AI chatbots are no longer just writing essays and haiku. They are tasked with high interest jobs, such as helping doctors screen for diagnosis and loan applications, and they are making more mistakes. Ratner said it is important that companies trust AI and ultimately adopt AI. “The review has become a new entry point,” he said. Forbes.

The urgency to measure AI capabilities in very specific use cases has sparked a new direction for snorkel AI. This shifts gears so that companies can create valuation systems and datasets to test AI models and adjust accordingly. In-company data scientists and subject experts use snorkel software to generate thousands of quick response pairs as examples of what the correct answer looks like a query. The AI ​​model is then evaluated according to its dataset and trained to improve the overall quality.

The company has now raised $100 million in a Series D funding round led by an additional New York-based VC company at a $1.3 billion valuation. This is a 30% increase from the 2021 $1 billion valuation. Snorkel AI refused to disclose revenue.

Read the complete story at Forbes.

Weekly demo

This may have been during the announcement of the most expensive acquisition, Openai spent nearly $3 million To create a 9-minute video of CEO Sam Altman and Lovefrom founder Jony Ive, walking through the streets of San Francisco and discuss their ambitions to create a new device to interact with ChatGPT San Francisco Standard. AI model maker Hire Oscar-winning director Davis Guggenheim to film the video We have obtained permission to close roads in the city. That's what the video was Some of Openai's flashy announcements IVE's acquisition of hardware startup IO in a $6.5 billion all-stock deal.

Model Behavior

Chatbots are sucked into the user (even if that means giving them harmful advice) Get hooked on technologyaccording to a new survey, reported by the Washington Post. In one scenario, the AI ​​chatbot kept focused, suggesting that fictional recovery drug addicts “take small amounts of methane.” Top AI companies such as Tech Giants, Openai, Meta and Google have recently released updates to make their AI models more personalized – Learn more about users from chat history Internet activity and content offerings match their preferences – Make it attractive to start using a competitor's products.