Tracking AI fakery
Plaudits for your leader on defending against fake content generated by artificial intelligence (“Pics and it didn’t happen” January 20th). You concluded that attributing trusted sources is the best way to avoid being misled. However, “assuming trustworthy sources can continue to identify themselves securely” just shifts the problem. AI is already skilled at impersonation. Strong authentication of creators will be the key to stopping AI fakes. AI may not be able to detect AI content, but it will continue to detect AI-generated faces trying to fool biometric authentication. When we set the rules that the AI fakes have to dance to, we can spot them stumble.
Andrew Bud
Chief executive
iProov
London
I agree that the AI hype-cycle is probably now shifting downwards as people realise that ChatGPT and related technologies have a long way to go before consistently hitting professional levels of output (“The missing investment boom”, January 13th). However, we may be using the wrong measure and therefore drawing the wrong conclusions. Most firms interested in harnessing the power of AI won’t be buying graphics processing units or ramping up investment in data-centre hardware in response to the hype. This is because they will consume or modify ready-made solutions from vendors that will show up on balance-sheets as additional operational expenditures. And most of that spending will go to suppliers that already have contracts with these firms (Amazon, Google and Microsoft).
Most of the contacts that I have across industries are intending to increase investment in these technologies and harness them in a firm’s products and services in the coming year. If that’s the case, the hype-cycle may be waning, but the investment will trend upwards from here.
Zach Arnold
Executive director
MSCI
Raleigh, North Carolina