Welcome back to The Prompt.

On Monday, OpenAI announced it had fixed a bug in ChatGPT that made it appear as though the chatbot was sending messages to users without prompting. ChatGPT proactively asked a user, “How was your first week of high school?” The person posted a bizarre exchange on Reddit asking the AI ​​chatbot, “Did you just message me first?” The company claims the bug occurred because a previous message was not delivered properly.

Plus, want to know about the latest technologies before they go mainstream? Sign up for our new weekly newsletter, The Prototype. The first issue is out this Friday!

Now let’s get to the headlines.

AI OFFER OF THE WEEK

Salesforce Ventures, the investment arm of the software giant, is planning another 500 million US dollars to generative AI companies, bringing the total funding provided to AI startups in the last 18 months to 1 billion US dollars, Forbes Through the company’s generative AI fund, it has supported leading startups in this field, such as Anthropic, Coherence, Mistral and Hugging Face.

Also noteworthy: World Labs, a startup co-founded by Stanford scientist and “godmother of AI” Fei Fei Li, has $230 million raised by Andreessen Horowitz, NEA and Radical Ventures to build spatial intelligence through AI models that can understand and interact with objects and places in the physical world.

BIG GAMES

OpenAI has released a new family of large language models called OpenAI o1. They are designed to take more time than GPT models before responding to queries, so that they work out different ways to answer a question and provide better, more accurate answers. They are built to “Reason” about complex problemsAccepting and rejecting options when editing a task.

Nikunj Handa, who works on OpenAI’s API team, tells me that the intelligence of the model is comparable to that of a “extremely clever doctor” and can perform advanced tasks in Law, Coding and Science fields. This is done through the use of so-called “Chain of Thought” Argumentationa step-by-step process of problem-solving. “People go to great lengths to get an LLM to do what they want. And this model makes that easier, almost by design, without having to tell it, ‘Please don’t output things in this format,'” he said.

These models are in the early stages and are not connected to the web or other external tools, and their access is limited to developers who have spent at least $1000 on the platform. There are also Security concerns. When testing earlier versions of the models, model evaluation firm Apollo Research found that the models capable of deceiving its users by giving the impression of being safe and helpful, while in reality they do not respond appropriately to requests.

DATA DILEMMAS

AI companies have used data from Mumsnet, a UK-based Parents Forum with an archive of more than six billion Words, Wired reported. The platform has since attempted to conclude licensing agreements with AI companies like OpenAI to sell its mostly female-authored content. But talks with OpenAI, which initially expressed interest in the site’s data, fell through after the AI ​​giant said it was entering into partnerships for large data sets that are not publicly available. Now Mumsnet is taking legal action against the company and other scrapers because alleged copyright infringement.

DEEP INSIGHTS

Black-and-white portraits of Rosa Parks, letters from Thomas Jefferson and the Mainz Giant Bible, a 15th-century manuscript believed to be one of the last handwritten Bibles in Europe, are among the 180 million items, including books, manuscripts, maps and audio recordings, housed at the Library of Congress.

Every year, hundreds of thousands of visitors stream through the library’s soaring colonnades, passing under Renaissance-style domes adorned with murals and mosaics. But recently, the more than 200-year-old library has attracted a new kind of customer: AI companies eager to access the library’s digital archives—and the 185 petabytes of data stored within them—to develop and train their most advanced AI models.

“We know we have a large amount of digital material that major language modeling companies are very interested in,” said Judith Conklin, Chief Information Officer of the Library of Congress (LOC), Forbes“It is extremely popular.”

But those who want to access the library’s data must access it through the API, a portal that allows anyone from genealogists to AI researchers to download data. They are prohibited from scraping content directly from the site, a common practice among AI companies that Conklin says has become a real “hurdle” for the library because it slows down public access to its archives.

“Others want our data to train their own models, but they want it fast, so they just crawl our sites,” she said. “If they impact the performance of our sites, we have to manually slow them down.”

The hunt for data is only part of the story. Companies like OpenAI, Amazon and Microsoft are also vying for the world’s largest library as customer. They claim that AI models can help librarians and subject matter specialists with tasks such as navigating catalogues, finding records and summarising long documents. This is certainly possible, but there are some rough edges that need to be ironed out first. Natalie Smith, the LOC’s director of digital strategy, said Forbes that AI models trained on contemporary data sometimes have problems with historical accuracy – for example, they identify a person holding a book as someone holding a cellphone. “There is an overwhelming bias towards the present day and so they often apply modern concepts to historical documents,” Smith said.

Read the full story on Forbes.

WEEKLY DEMO

billionaire Venture capitalist Walter Kortschakwho has invested in technology companies such as Twitter and Lyft for more than four decades, said Forbes that it is dangerous swept away in The “mass hysteria” of AI. The investor, who owns shares in OpenAI and Anthropic, offered some advice: it is important to be selective, “invest in founders who are outsiders” and discipline the number of investments made.

QUIZ

OpenAI’s new o1 model was able to correctly answer questions from this difficult TV quiz show:

1. Danger
2. The floor
3. Just connect
4. Wheel of Fortune

Check here if you have understood it correctly.

MODEL BEHAVIOR

Snapchat has the right to use AI-generated images from users and their friends in Advertisementaccording to the terms of service of the social media company’s “My Selfies” tool, which lets users upload their photos to create AI-generated images, 404 Media reported. Although the feature can be disabled, it is enabled by default. At least one person came across an ad in Snapchat stories that showed an AI-generated image of themselves, the person posted on Reddit.