Today:

ChatGPT custom instructions are POWERFUL Replace AutoGPT and BabyAGI?

OpenAI has launched a new feature called "Custom Instructions," and it's pretty awesome. You can now tailor your interactions with ChatGPT, making it more personalized and efficient. You can opt-in by enabling the feature in your profile settings. By providing specific instructions, you can significantly impact how ChatGPT responds to your prompts.

The real power of Custom Instructions lies in creating a memory system for ChatGPT. By saving and retrieving information from a text file, ChatGPT can access past conversations and use them to improve its reasoning and planning abilities. It's like building a memory bank for the AI.

You can do cool stuff with this feature, like creating a Keto Plan, developing a high-protein meal plan, or even writing tweets in someone else's style. The possibilities are vast, and people are already coming up with exciting ideas for AI agents.

WATCH THE VIDEO ON YOUTUBE

Top AI companies visit the White House to make ‘voluntary’ safety commitments

The White House has been playing catch-up with the rapid pace of AI development. In response, the Biden administration roped in seven top AI companies, including OpenAI, Google, Microsoft, Meta, and Amazon, for a voluntary agreement to prioritize safety and transparency.

But let's get real here - no rules or enforcement, just a gentleman's agreement. However, it's likely gonna be public knowledge if a company doesn't hold up their end of the bargain.

Big shots from these companies paid President Biden a visit to discuss this deal, committing to a bunch of measures: security tests of AI systems before release, sharing info about AI risks, investing in cybersecurity, establishing third-party bug discovery, marking AI-generated content, reporting AI systems’ capacities and limitations, and prioritizing research on societal risks. They also aim to use AI to tackle big societal issues like cancer and climate change.

READ THE ARTICLE ON TECHCRUNCH

Cerebras Systems signs $100 mln AI supercomputer deal with UAE's G42

Tech company Cerebras Systems has got a fat $100 million deal in its pocket, after teaming up with the United Arab Emirates tech group, G42. They're going to deliver the first of potentially nine AI supercomputers.

Cerebras, which hails from Silicon Valley, shared that G42 has agreed to buy three of its "Condor Galaxy" systems, which will be built right here in the U.S. The first machine will be up and running this year, and we can expect two more by early 2024.

The top man at Cerebras, Andrew Feldman, is calling this deal the beginning of a "strategic partnership." They're already in talks to provide up to six more supercomputers by late 2024. Feldman is packing his bags to spend three months in the UAE to work with G42, calling it "a rare opportunity to change a massive market."

G42, based in Abu Dhabi and made up of nine operating companies, plans to use the Cerebras systems to sell AI computing services to healthcare and energy companies. They've got $800 million backing from U.S. tech investment firm Silver Lake, which in turn is backed by the UAE's sovereign wealth fund, Mudabala.

READ THE ARTICLE ON REUTERS

OpenAI’s head of trust and safety Dave Willner steps down

OpenAI’s Head of Trust and Safety, Dave Willner, has stepped down to transition into an advisory role and spend more time with his family. He was in the role for about a year and a half. In the interim, CTO Mira Murati will manage the team.

This announcement comes at a critical time in the world of AI. The company recently committed to shared safety and transparency goals in a White House meeting. There are numerous questions surrounding how best to regulate and control the potential harmful impacts of AI, and Willner’s role was central to these discussions.

Willner is a veteran in the industry, having led trust and safety teams at Facebook and Airbnb. He was an early Facebook employee who helped define the company's first community standards position, which is still in use today. His work at OpenAI focused on ensuring technologies like the image generator DALL-E were not misused.

READ THE ARTICLE ON TECHCRUNCH

Microsoft Teams is adding Maybelline’s AI-powered ‘makeup’ filters

If you've ever dreaded turning on your webcam during a Teams call because you're feeling a little under the weather, Microsoft's got your back. They've partnered with cosmetics giant Maybelline to bring a new AI-powered beauty feature to Teams.

This new feature will give you access to 12 unique "virtual makeup" filters that can help you shine even on your off days. These filters mimic the effects you often see on social media platforms like TikTok. You'll be able to play around with various blurring effects and digital makeup color options. And if you love the look? The app will give you a breakdown of the actual Maybelline products used, so you can bring the virtual into the real world.

The tech behind these fancy filters comes from Modiface, an augmented reality company specializing in the beauty industry. They're a popular choice for virtual makeup experiences and have been used by big-name companies like Sephora, Estée Lauder, and L'Oreal, which also happens to be Maybelline's parent company.

The new feature is rolling out to global Microsoft Teams Enterprise customers. You can find it under the "Video Effects" tab in the Teams meeting settings. It's a bit of a head-scratcher to see beauty filters in a professional setting, considering the mental health concerns tied to body image that they can trigger. But clearly, folks are keen on looking their best, even in work meetings, as seen in other video conferencing platforms like Zoom that also offer beauty effects.

READ THE ARTICLE ON THE VERGE

CodeSee adds generative AI to explore code bases with natural language queries

CodeSee, a company that helps programmers understand complex code bases within a company, is introducing generative AI capabilities. This new feature makes it easier to ask questions about the code in natural language. With the AI's help, the system can provide text-based answers, and also show a visual map of the code base to demonstrate the connections related to the question.

CodeSee's CEO, Shanea Leven, believes this AI addition will enhance how people comprehend and navigate code. With this, programmers can ask about specific code functionalities, and CodeSee will show where in the code these functions are found, providing a clearer understanding of how things work.

The company worked with Microsoft and OpenAI to develop this AI feature, which added advanced capabilities without the need for additional employees. This addition will augment CodeSee's existing service maps that offer visibility across multiple code bases, helping enterprises better understand their coding systems.

READ THE ARTICLE ON TECHCRUNCH

AI In Banking, Payments and Insurance

AI's been making big waves in banking, payments, and insurance. Algorithms are behind the scenes, making decisions about our finances every day. Right now, AI mostly helps with marketing and customer service, where chatbots handle the basics, freeing up folks to tackle tougher issues. But with generative AI like ChatGPT in the mix, even complex queries could be handled by the bots.

Businesses are tiptoeing around AI use. They're still trying to figure out how to best use the tech. Simon Lyons from The London Institute of Banking & Finance reckons AI's real worth is in spotting trends and making sense of them. AI's got a big role in crunching numbers, processing data, and tackling fraud. Ellen Logan from Augmentum Fintech explains that predictive AI helps make rules-based decisions for stuff like underwriting, fraud detection, and trading strategies.

In insurance, AI's got its mitts on claims assessment, pricing, and personalizing cover. But there's concern about AI being used for individualized pricing. Consumer group Fairer Finance wants boundaries and transparency on this.

READ THE ARTICLE ON FINANCIAL TIMES