The SEC Newgate AI Weekly

AI at the Superbowl 

The Superbowl took place last Sunday, generating 123 million viewers, making it the USA’s ‘most watched show ever’. Alongside the slow but inevitable march of the Patrick Mahomes-powered Chiefs to victory, the ads (as usual) were a significant talking point, and a number of them were AI-generated. Microsoft chose to promote their product Copilot, while Google chose to promote their Guided Frame feature, which features voice prompts to guide people with limited sight on when faces are in frame for a selfie. We also saw ads from CrowdStrike, Etsy, Despicable Me 4 and Avocados from Mexico. What really stands out is that each ad is for an AI-led product, yet they all promote very different uses of AI. You can view the ads here. 

Sam Altman calls for independent regulation

The CEO of OpenAI, Sam Altman, has made another warning about the dangers of AI. As AI is advancing faster than expected, he has called for a regulatory body like the International Atomic Energy Agency to oversee AI. He argues that the AI industry itself should not be in charge of making regulatory decisions and there is a growing need for an independent regulatory body.

ChatGPT is getting a memory

ChatGPT is continuing to evolve, with the latest feature being the ability to remember personal details from past conversations, allowing it to apply those memories to current queries. This will certainly help to deliver a more personalised experience for each user, with GPTs that are tailored specifically for users’ needs. You will still be able to adjust what it does and does not remember.

Tech giants working together to combat political deepfakes 

One of my predictions for 2024 was that there would be a ‘race to the bottom’ from politically motivated operators to deliver political deepfakes into people’s newsfeeds. So it is encouraging that Meta, Microsoft, Google and OpenAI are currently working on an agreement to stop political deepfakes intended to deceive voters. With crucial elections coming up this year, including in both the US and the UK, there is expected to be an accord announced at the Munich Security Conference on Friday. The accord will look at ways to identify, label and control misleading AI-generated images. It is yet to be seen how far a watermark can go in convincing someone an image is fake…