What is meant by asking what technology holds the greatest promise or where the best trends lie? Are we going to argue about it? Of course not. What we can actually identify now as a super-trend is the so‑called 3M, or M cubed. First of all, this of course refers to mega-transformers. <...> It’s about multimodality, when different modalities are processed by this model. By this, I mean it’s something that can read and write. <...> It can tackle both financial and non-financial tasks. It’s really amazing stuff. In terms of multitasking, gate to deep mind, for example, 600 tasks can be handled simultaneously — Alexander Vedyakhin, First Deputy Chairman of the Executive Board, Sberbank.
When it comes to tackling our tasks and what we use for this, it cannot be about one issue alone. Given the way all issues surrounding artificial intelligence are unfolding at such a rate, they need to be separated into a minimum of two branches. The first covers technology and production. Why is that? Because implementation has begun apace. Industrial-scale implementation requires us to review how we work with technology. <...> The second branch is about pushing this intelligence to a new level. In this regard I’m in absolute agreement with my colleague — Anton Fedchin, Head, Odnoklassniki.
It has now begun to turn into an exact science. <...> If the model – and this is already a precise statement – does not change too much during the learning phase, then we can say virtually everything [about it – ed.]. That includes transformers. Basically, the size of the model doesn’t matter. This is a multidimensional problem, and one very effective way to tackle tasks of this kind is through approximation — Alexander Kuleshov, President, Skolkovo Insitute of Science and Technology (Skoltech).