The SEC Newgate AI Weekly

Realignment? OpenAI’s risk team exits

A bad week for anyone concerned about AI ethics – following the departure of OpenAI co-founder & Chief Scientist Ilya Sutskever and his superalignment co-lead, Jan Leike, Wired got the scoop that the team, tasked with overseeing the long-term risks of AI, has been disbanded.

OpenAI is believed to operate an extremely strict NDA policy which means we’re unlikely to hear more from those leaving, but it leaves big doubts about the company’s approach to safety.

Gen Z’s Gen AI concerns

Buried away in the Deloitte 2024 Gen Z and Millennial Survey is the news that 59% of Gen Z and 52% of millenials say that “the prevalence of Generative AI wil make them look for job opportunities that are less vulnerable to automation”. Those numbers rise to 78% and 79% respectively for frequent users. One to think about for businesses seeking young talent in the wide range of fields vulnerable to automation.

I Reddit on ChatGPT so it must be true

We’ve been talking for a while now about the reputational risks of generative AI to business and individuals. The partnership announced by ChatGPT and Reddit further increases those risks, with the chatbot now able to reference Reddit as a source in answers to questions. If you’re not monitoring Reddit for misinformation about your company, it just became a whole lot more important to start immediately.

Are you having a laugh?

Great news from the Netherlands where researchers have built an AI-driven sarcasm detector. The AI, which has been trained on sitcoms such as Friends and The Big Bang Theory, could, for example, vastly improve the ability of social listening tools to determine sentiment accurately and make chatbots more able to understand the nuances of human conversation. We can’t wait.

OpenAI to get a bite of the Apple?

Rumours are swirling that Apple will announce a partnership with OpenAI at its WWDC conference in June. The company appears to be behind the curve in developing its own generative AI technology so it makes sense to partner whilst they catch up. If true, generative AI chatbot usage could go from millions to billions with a single update to iOS. Very much one to watch.

Product Recall?

Microsoft have run into trouble with privacy campaigners and regulators following the announcement of Copilot PCs – personal computers built with AI in mind rather than as an add-on. At a special event on 20th May, Microsoft unveiled a range of features but the one that caught the eye of the UK’s Information Commissioner’s Office (ICO) was Recall, which allows users to look back over their device usage to find specific actions they have undertaken in the past. The example Microsoft used in their demo was of someone looking back over months of browsing on an ecommerce platform to locate a previous search for trainers. But the ICO has asked for more information on the technology behind it, which involves taking ‘snapshots’ every few seconds, to determine whether the company has put sufficient privacy safeguards in place. Microsoft says it has, and has mentioned opt-outs as a possible compromise. But how useful will it be if some tasks are recorded and some aren’t?

And finally...

With the UK general election now underway, a lesson from the US for any political strategists considering the use of deepfake content. The political consultant behind the Joe Biden deepfake robocall has been indicted. Voters received automated calls using the President's voice which told them to stay at home in January's New Hampshire primary and save their efforts for the Presidential election in November. This was thought to be the first recorded deepfake to be deployed in a major US election. Whilst we expect to see something of this nature during the UK election campaign, the indictment will hopefully make anyone considering the tactic think twice.