Last Month in AI:
The Wait

Instead, OpenAI continued to drip-feed AI tech to the world, releasing just enough to keep its GPT-4 model at the top of AI leaderboards. To this end, GPT-4 now has an upgraded memory, so it can remember when you’ve been nasty or nice to it. Cute. OpenAI’s backer, Microsoft, also played the drip-feeding-AI-to-the-masses game by showcasing, but not releasing, its VASA-1 deepfake tech that turns a single photo of you into films of you doing anything they want you to be seen doing. Hmm.

The logic of the drip-feed-the-public approach fits OpenAI’s remit to progressively reveal to public for what AI can already do, without creating mass panic. It seems to be working; the Future of Humanity Institute for AI worriers shut itself down in April, perhaps indicating the robots aren’t coming for us after all. But what about Q*?

Despite The Wait, the AI world didn’t stand still in April 2024. Microsoft and Coca-Cola partnered to accelerate generative AI adoption, WPP and Google teamed up to reinvent AI generated marketing, and Accenture announced it was on track to deliver $2.4 billion in generative AI earnings. Anthropic’s new Claude 3 AI stunned researchers with impressive self-awareness, whilst Meta launched an ‘open-source-ish’ Llama 3 model, and a new Meta.ai assistant to rival GPT-4. Meanwhile, Udio‘s new AI music generator impressed, as did Professor Ethan Mollick’s new AI survival manual “Co-intelligence“.

Celebrity AI stories filled the gap left by a GPT-5 gone MIA. Notably, rapper Drake used AI to diss fellow rapper Kendrick Lamar in a track that used the cloned voices of Snoop Dogg and the deceased Tupac. That got him a cease-and-desist order while giving audiences an idea of what weaponizing AI means in 2024. Deepfakes, baby.

There was something for everyone in April to keep themselves occupied during The Wait. The ‘AI Pessimists’ found their fears realized when it was reported that AI was already being used to frame suspects, create non-consensual deepfake nudes, displace jobs, and generally freak out people working in cognitive and creative industries. They worried about news that a third of translators and a quarter of illustrators are losing work to AI, with journalists and essay graders next in line for obsolescence.

Meanwhile, the ‘AI Optimists’ found hope during The Wait, with AI being put to use for gene editing, predicting sepsis, fighting conspiracies, fact-checking, and boosting productivity by 14%. The mass rollout of AI assistants in business and business schools continued, with an emerging norm, if not right, for every student and every worker to have their own AI assistant.

But it was the ‘AI Sceptics’ who were happiest about The Wait in April, which they predict will be long, if not unending. They relished the dire reviews of much-hyped AI devices, Humane’s ‘Pin’ and the Rabbit R-1 from Rabbit Tech. They were delighted that Amazon announced it was dead-poling its AI-powered ‘just walk out’ checkout-less tech, which turned out to be powered by hundreds of Amazon workers watching shoppers on CCTV rather than AI itself. And no doubt they revelled in the fact that the smartest AI on the planet, GPT-4, is still so dumb that it can’t list post-war German chancellors with 2 ‘e’s in their name (try it).

The Great Wait may or may not continue. But even if it does, today’s GPT-4 level tech can do astounding things. Most notably, it can level up and upgrade those of us who are new, less skilled, or less talented. To this end, and without any musical talent whatsoever, I worked with ChatGPT to turn an insightful article by Wharton’s Ethan Mollick on thinking about AI assistants as real people into the lyrics of a song. I then popped this into the AI music generator Suno, and the result is what I think is a rather wonderful and thought-provoking track.

Who needs GPT-5 when we can already make magic with AI?

Song: In a world of code and neon dreams