Continuing Innovation – AI’s Next Wave

Now that most people know a generative AI and the work of neural and LLM networks, what is the next step? What will the market are going to live and that the industry will advance?

There was a panel at our IIA event in Davos in January which fought with some of these questions. It included a fairly good range of commercial representatives with Google, TCS and others in the mix.

Moderate by Daniela Rus, my colleague from MIT CSAIL LAB, these experts have exceeded some of the main AI trends which are likely to influence its use in the coming years.

Failure of laws and architectures

Most panelists have agreed that ladder laws are still alive and always relevant.

“People have crossed all the internet sections and have nestled as many words and corpus as possible, and hit this wall a little in terms of training you can use,” said Panelist Jack Hidary, CEO of Sandboxaq. “And of course, you must spend billions of dollars … in other areas of AI, however, and when we generate data, synthetic data from equations or sensors or other types of sources, we don’t even have to hit a data wall. So I think we can evolve in a very spectacular way. »»

However, there is a passage from models of transformers to other new types of versions for LLMs which will make them more distributable, more decentralized and capable of operating on Edge devices.

For example, Mathias Lechner, CTO of Liquid Ai, has exceeded certain aspects of liquid networks which are smaller and thinner constructions, and the savings of resources attached to it. I wrote a lot on liquid networks, so you can go back and look at the technical details. In addition, this is the common warning: that I have consulted the liquid AI and that I know the people also working on this technology at CSAIL.

“The cost of inference, the cost to operate these models, you know, must be taken into account,” he said. “It is important: forming such a huge model, you know, run and make this model (more expensive) work compared to a much more efficient model, which is much more optimized.”

In the future of applications, panelists have suggested that we should examine engineering solutions.

“We will see more efficiency,” said Peter Sarlin de Silo. “We will see a kind of improved performance. At the same time, I think that we are always somehow focusing on AI assistance for general use and very specific use cases, as has been highlighted, and … we have a lot of new modalities that must be addressed, and there is a huge value to that. »»

Bilingual talent

What about people who need to inaugurate the next generation of AI?

We may not need linguistic experts, but we will need people with very special skills sets.

Hidaire has spoken of the juxtaposition of chemistry, biology and physics, and suggested that with quantum computer science thrown into the mixture, there will be a need for bilingual professionals who are trained not only to AI, but also to other disciplines such as chemistry.

He also mentioned how much the quantitative AI is executed on GPUs, which is an important development in the material space.

Put products on the market

The panel also explained how to generate income with AI.

Harrick Vin said TCS is considering a process of “increasing people” through digital processes to improve the quality of their decisions. This is something, he said, it’s underway now.

Hidary spoke of biopharmate and chemistry, and the use of carbon dioxide for practical purposes.

Others have discussed more generalized use, where AI underpins all kinds of capacities on different verticals.

The biggest context

It is clear that we are moving to new models that will contribute to optimization and efficiency.

One thing that has not occurred as much in this panel is the movement of work. In an article that I did recently, I examined various types of common jobs that could soon be taken up by AI entities. It is also obvious that agency AI is the path of the future. But this panel explained how companies will use these capacities to their advantage, and it is a dignified subject, to understand what things will look like relatively soon.