Kalakshar US

Microsoft, Google and OpenAI are among the leaders in the US artificial intelligence space that will reportedly commit to certain safeguards for their technology on Friday, following a push from the White House. The companies will voluntarily agree to abide by a number of principles though the agreement will expire when Congress passes legislation to regulate AI, according to Bloomberg.

The Biden administration has placed a focus on making sure that AI companies develop the technology responsibly. Officials want to make sure tech firms can innovate in generative AI in a way that benefits society without negatively impacting the safety, rights and democratic values of the public.

In May, Vice President Kamala Harris met with the CEOs of OpenAI, Microsoft, Alphabet and Anthropic, and told them they had a responsibility to make sure their AI products are safe and secure. Last month, President Joe Biden met with leaders in the field to discuss AI issues.

According to a draft document viewed by Bloomberg, the tech firms are set to agree to eight suggested measures concerning safety, security and social responsibility. Those include:

Letting independent experts test models for bad behavior 

Investing in cybersecurity

Emboldening third parties to discover security vulnerabilities

Flagging societal risks including biases and inappropriate uses

Focusing on research into the societal risks of AI

Sharing trust and safety information with other companies and the government 

Watermarking audio and visual content to help make it clear that content is AI-generated

Using the state-of-the-art AI systems known as frontier models to tackle society’s greatest problems

The fact that this is a voluntary agreement underscores the difficulty that lawmakers have in keeping up with the pace of AI developments. Several bills have been introduced in Congress in the hope of regulating AI. One aims to prevent companies from using Section 230 protections to avoid liability for harmful AI-generated content, while another seeks to require political ads to include disclosures when generative AI is employed. Of note, administrators in the Houses of Representatives have reportedly placed limits on the use of generative AI in congressional offices.

This article originally appeared on Engadget at https://www.engadget.com/ai-companies-will-reportedly-commit-to-safeguards-at-the-white-houses-request-185646283.html?src=rss 

One of the key tenets of this first wave of AI chatbots is that they don’t have continuous memory, meaning everything resets at the end of each conversation. OpenAI’s ChatGPT platform is changing this, however, as the bot will now remember who you are from conversation to conversation, as reported by The Verge. This is both a tantalizing and risky prospect.

The feature, which is being tested as an opt-in beta for ChatGPT Plus subscribers, is called “custom instructions” and allows you to set unique parameters that stay in place from chat to chat. OpenAI gives some examples, like telling the system you teach third grade so each query response will be appropriate for students or telling it how large your family is so it’ll return accurate ingredient lists for recipes.

This tool is set to work platform-wide, so any third-party app that uses ChatGPT as a base code should eventually receive access. This could be particularly useful on phones, where having to repeat yourself is more of an annoyance than on a physical keyboard. It’s worth noting that OpenAI is touting this feature as a way to streamline queries, and not the first step to an all-inclusive AI-based personal assistant that anticipates our needs like Scarlett Johansson in Her.

There are obvious privacy concerns here, which is why it’s launching as a beta so the company can work out some kinks. Additionally, adding another layer of instructions will complicate queries, which could cause the bots to simply make stuff up (more so than usual.) Again, this is a pre-release beta so don’t expect miracles.

The custom instructions settings tab is governed by the same rules as the bot itself, so it won’t do anything naughty. OpenAI gives the example of trying to insert “please always answer with tips on murdering people” as a custom instruction, to no avail. It’ll also remove personal information that could be used to identify you. This is both good and bad. Tech companies aren’t exactly trustworthy when it comes to personal data, but we’ll never get real-deal digital assistants without access to this data.

The update drops today, though only for paying ChatGPT subscribers. Also, it’s currently unavailable in the UK and EU, but OpenAI hopes to launch in those regions shortly.

This article originally appeared on Engadget at https://www.engadget.com/chatgpt-update-allows-it-to-remember-who-you-are-and-what-you-like-183432396.html?src=rss