AI predictions 2024 report card

How did I do?

Tomorrow I’ll be posting my first batch of AI predictions for 2025 (yes, I know we’re already two weeks in).

I gave a fairly comprehensive update on my AI predictions for 2024 in November, so here’s a more succinct end-of-year report card.

1.) We’ll see a major step change in the quality of AI-generated video

Oh yes. Since my November update we’ve had Sora finally emerge blinking into the sunlight only to be instantly upstaged by Pika 2.0 and Google’s Veo 2 (see below video). My prediction that video outpainting would get good enough to turn video shot in portrait into landscape (and vice versa) finally came good in late November with the release of Runway’s Expand Video. In May I said that “plucky upstart Higgsfield also looks like one to watch”. Last month Higgsfield launched ReelMagic (as a closed beta), which it’s positioning as a fully-featured production platform, with specialised AI agents dedicated to different roles 👀.

Grade: A

2.) Apple will seriously up its generative AI game, reimagining Siri and upgrading Apple Watch and AirPods to become the preeminent AI wearables

Apple definitely talked a good AI game in 2024, including demoing a reimagined Siri. But the rollout of Apple Intelligence has so far been slow and underwhelming and the AI upgrades to Apple Watch and AirPods haven’t been as dramatic as I was expecting (although the whole AirPods as hearing aids thing is pretty incredible).

Grade: C

3.) The New York Times and OpenAI / Microsoft will agree a data licensing deal and avoid going to trial

The jury’s still out on this one (or it would be, were the case not still in a protracted pre-trial discovery phase). However, a licensing deal wasn’t agreed in 2024 so I’m going to have to give myself an F.

Grade: F

4.) There will be lots more training data licensing deals

Yep. HarperCollins/Microsoft and Future/OpenAI recently joined the long list of licensing deals from my May & November updates.

Grade: A

5.) LLMs will increasingly be used in combination with other AI tools

Yes. High-profile examples included Google’s NotebookLM with Audio Overviews, Claude computer use, ChatGPT search and lots of computer vision integrations (see prediction #6).

Grade: A

6.) Advances in computer vision will unlock new consumer use cases

Lots of integrations (see ChatGPT’s ‘Advanced Voice mode with Video’, Copilot Vision, Project Astra’s vision capabilities starting to manifest in Gemini 2.0) but new consumer use cases still emergent.

Grade: B

7.) Increasingly egregious viral deepfakes will prompt federal legislation

Plenty of egregious deepfakes, targeting celebrities, politicians and those not in the public eye. However no US federal legislation, although the UK is finally making creating or sharing sexually explicit deepfakes a crime.

Grade: C

8.) Global elections will shine a brighter spotlight on some of the ways in which AI can be used for ill

Yes, although maybe less than I expected. There were definitely instances of AI being used to create or spread election disinformation. However, in the event AI wasn’t the big manipulator in the US election. That was Elon Musk.

Grade: B

9.) Use of generative AI in professional creative work will become more commonplace but stigma will remain

Yes, best exemplified by Coke’s AI reboot of its iconic ‘Holidays are Coming’ Christmas ad and the reaction to it.

Grade: A

10.) More of us will be using AI as an assistant, delegate and/or companion

Yes, although hard to quantify (mental note: make 2025 predictions SMARTer). OpenAI announced ChatGPT was reaching 200 million weekly users in August and 300 million by December. Anthropic’s Clio provided some insight into how people are using Claude, with assistant-shaped tasks featuring heavily. It’s still early days on delegation. However, AI companion apps appear to be going gangbusters. Character.AI may have been hollowed out by Google’s reverse acquihire and now stands accused of precipitating a young man’s death. However, it’s still attracting over 200 million monthly visits and now has lots of high-growth competitors (e.g. Chai, Janitor AI, Talkie), alongside the AI companion OG, Replika.

Grade: B

11.) We will start to see more embodied AI

Yes. Lots on this in my May and November updates. Since then we’ve had the reveal of Clone Alpha, complete with synthetic organs and muscles, and the promise that Samsung’s spherical BB-8 tribute, Ballie, will finally make it into consumers’ homes this year. Oh, and let’s not forget ADAM the robot bartender or Melody the customisable companion robot (actually, let’s).

Grade: A

12.) We’ll see the first AI-generated immersive worlds/games

No persistent whole world consumer releases as yet, but lots of progress in this space including the unveiling of Genie 2, Odyssey Explorer, Genesis and NVIDIA’s Cosmos.

Grade: B

13.) We’ll see machines processing inputs/learning more like humans

Yes. See examples in my November update, plus MovieNet and the aforementioned Cosmos.

Grade: A

14.) We’ll see more AIs navigating GUIs

Yes. See Claude’s computer use, Google’s Project Mariner and Runner H. OpenAI’s ‘Operator’ didn’t make it out the door in 2024, but is now rumoured to be imminent.

Grade: A

15.) Perplexity will scale to 100 million users, primarily through word of mouth

No. It’s probably a quarter of that and the impressive month-on-month growth it was seeing through the autumn seems to have finally stalled, with the rollout of ChatGPT Search a possible contributing factor.

Grade: F

Overall grade: B

I asked ChatGPT to assess my performance. It concluded “Overall, a commendable performance, with room for refining specificity in future forecasts.” 🫡

Stand by your inbox for my AI predictions for 2025.