2024 Predictions

Large Language Models

Google Ultra is due to be released in the first part of this year. I predict that behind the hype and the eagre marketing there is some solid product, but that it won’t exceed the capabilities of chatGPT4. Which 9 months after release is still the market leading AI, recently valued at $100bn.

I expect that we will see Claude 2.5, ChatGPT 4.5 (or 4.7 depending on how you count). We probably won’t see the release of combined google search experience and bard in general replacement on Google.com. The commercial impact on the Google ad revenue is still broadly unknown. In the open source chat bot (LLM) space, expect an update to LLAMA, and more solutions from Mistral.

The biggest challenges which I foresee being resolved in 2024 are the context window (and its issues even on large context) and much greater reduction in hallucination rates.

Image generation

The rate of progress on image generation will continue to be frenetic. This next year will almost certainly see solutions that cover, hidden watermarks (which no consumer really cares about) will come and go. Greater resolution and quality, more improvements to bias removal and continued growth of natural language, as opposed to contrived prompts for image generation. Text, fingers, consistent images and people will probably be in the most part solved, or have viable workarounds.

Most stock libraries will either be licensed into training sets, or they will have their own ‘generator’. Adobe is likely to consolidate the professional use-cases with continued generative tools, and workflow improvements for professional design.

Canva will continue to grow and fill in its AI capabilities for the production of pro/consumer grade graphic design output. It will be interesting to see how this will be impacted by co-pilots in powerpoint.

Video

Both video generation, and video editing will have a breakout year in 2024. Similar to the still image growth of 2023. Video is a much more compelling and viewed media on both social, and commercial channels. The combination of better editing tools (background removal, key frames) and generation length will lead to rapid changes in this tool set.

Hello Mr Smith… or agents

Autonomous agents, robotic process automation to take AI to the next levels of practical use will be a big growth area. The launch of the openAI GPT store, will be like the launch of the app store back in 2010. Expect a LOT of bad software and a few gems.

Audio

Speech, both transcription and translation are a much under discussed AI application. The ability to ‘clone’ voices or generate entirely synthetic voices has made great progress in 2023. Considering that second to visual moving images, the spoken language is the most consumed form of media. Expect continued enhancements in the area of audio and speech AI this year. Higher quality, greater emotional range and even less input audio to create the clone.

Final thoughts

The rate of product releases, research papers has not slowed. So if you consider the position on the ‘S’ curve for technology. It still looks as though we are accelerating. I would expect this to continue through 2024.

2024 from a business and AI perspective, most conversations have moved from ‘what is it’, to ‘how do we get started’. With economic pressure continuing all businesses will be continuing to find effective growth strategy without operational cost growth.