Artificial Intelligence in 2025

Does it make sense to make predictions about AI, when one new LLM release (just think at what happened in the last 2-3 days and how that changed everything) can change the rules of the games overnight? Yes it still does, and it can potentially influence what AI will answer about its own future, if this article becomes part of its training set... So what will 2025 bring us, AI-wise?

The success of Multimodal AI: that is easy because the signs are already here.

The success of Meta's Llama: for businesses that are serious about having their own privacy-compliant AI that handles tasks with an eye on confidentiality, as long as they trained it properly to avoid prompt injection. And unless you plan to have a gazillion users on your platform or to use Llama to train other AI models, Llama is indeed open source to you.

The success of Google Gemini, as the standard choice for US individual recreational users (= not willing to pay much for their light usage) who want a strong and versatile AI.

The success of Claude, if you are based in the EU and need to wait 6-12 months for other models to give you access to their latest version. It may sometimes refuse to carry on your prompts ("I don't feel comfortable doing that" or "Traffic overload at the moment, try later") but it is a solid model.

The success of ChatGPT as the go-to provider of AI solutions for commercial-application and animations. Despite its IP holder name, it is not open and not that accessible anymore to the free user who wants to fully benefit from it, but with money coming in, it can invest back in product development

No, there will be no conscious AI in 2025. Despite all the hype, the pointless papers that try to disprove the opposite, the robots that fake awareness with an operator behind the wall: AI is not conscious. It will never be, as long as it is a GPU-based traditional computing code. It may become conscious, for example through entanglement, once quantum computers are a widespread reality that does not require absolute zero temperature to operate (think optics and acoustics), but that's not in 2025. I could say it will happen by 2050, adding the usual 25 years safety buffer that protect wild predictions from being criticised.

Yes, very likely, there will be a substantial Chinese AI success in AI. From known players that will refocus on AI as they lose access to key markets for other services, or as ROI on the massive amount of funding that China has been putting in research both at homegrown and foreign institutions.

Have a great 2025 everyone!