OpenAI says mysterious chat histories resulted from account takeover

OpenAI has responded to claims of ChatGPT leaking chat histories, explaining the issue as a result of an account compromise rather than a systemic flaw in ChatGPT.

A user, Chase Whiteside, noticed chat histories that included sensitive information such as unpublished research papers and login credentials appearing in his account, which he accesses from Brooklyn, New York. The investigation by OpenAI traced unauthorized logins to Sri Lanka, suggesting an account takeover rather than a leak from ChatGPT itself.

Whiteside, who used a strong password, was skeptical of the compromise explanation. This incident highlights the lack of protective measures like two-factor authentication (2FA) and login tracking on OpenAI's platform, a standard for most major services. The event underscores the importance of removing personal details from queries made to AI services to prevent potential data exposure. This follows a history of internet services experiencing similar privacy issues due to system errors or configuration mishaps, emphasizing the ongoing challenges in securing user data online.

Humanoid Robot Startup Figure AI in Funding Talks With Microsoft, OpenAI

Figure AI Inc., a company working on developing humanlike robots, is reportedly in discussions to raise up to $500 million in a new funding round, potentially led by Microsoft Corp. and OpenAI. In the proposed funding scenario, Microsoft might contribute approximately $95 million, with OpenAI adding around $5 million, potentially valuing Figure at $1.9 billion before the investment.

This valuation could position Figure as a significant unicorn in the startup world. However, the funding round details are not yet final and could change. Figure's flagship project, Figure 01, is an AI-powered robot designed to perform hazardous tasks unsuitable for humans, aiming to address labor shortages. The company previously secured $70 million in a round led by Parkway Venture Capital. The AI robotics sector has seen considerable activity, with various startups, including 1X Technologies AS and Sanctuary AI, raising funds and developing their own humanoid robots.

Midjourney’s “Niji V6” Is Here — Next-Level AI-Generated Anime Images

Midjourney has released Niji V6, an updated version of its anime-specific AI model. This new model offers advancements in understanding diverse anime styles from detailed prompts, encourages the use of long, descriptive prompts for better visual outputs, and can translate complex descriptions into anime visuals more effectively.

A new feature, "Style Raw," allows for the exploration of unrefined, creative anime expressions. Notably, Niji V6 now supports rendering text directly on images, enhancing its functionality. To use Niji V6, paying Midjourney users need to select "Niji Model V6 [ALPHA]" in the settings menu of the Midjourney chatbot. Examples of the model's capabilities are showcased in the niji-showcase channel on Discord.

Neuralink team successfully embedded the first chip in a human brain

Elon musk: "Even if someone has never had vision ever, like they were born blind,...we believe that we can still restore vision"

Elon Musk has announced that his company Neuralink has successfully implanted a wireless brain chip, named Telepathy, into a human for the first time. The procedure took place on Sunday, and according to Musk, the subject is recovering well with promising initial results, including neuron spike detection.

The device is designed to enable the user to control phones, computers, or almost any linked device simply by thinking about them. This development is part of the broader field of brain-computer interface (BCI) systems, which several companies are exploring. These systems involve implanting microchips into the brain to wirelessly send signals that can interpret a person's intentions, particularly those related to movement.

The initial target users for Neuralink's device are individuals who have lost the use of their limbs due to debilitating neurological conditions. The company received regulatory approval in the U.S. last year to carry out this human trial, which is still in the early safety assessment stages. While the technology is promising, it is noted that it could be several years before it is available on a wide scale, as the device's long-term operation and effectiveness remain to be thoroughly evaluated.

Other stuff

• Hottest Job in Corporate America? The Executive in Charge of A.I.

• AI companies will need to start reporting their safety tests to the US government

• Building GPTs tip: Don't use knowledge files

• Whisper powered by Apple Neural Engine! 🔥

Maximize Your ChatGPT Experience: Mastering Auto Split and Summarize with Superpower ChatGPT

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.