Andrew Ng's team recently asked what people's highest hopes for artificial intelligence is for 2026. Given that the year 2025 is widely considered the year that AI really started being adopted by the industry, and with even Sam Altman saying that he isn't sure what direction AI would take, the responses to the question asked by Andrew's team showcased a small snippet of what people expect to see.
I was honored that one published response was mine.
1. Inspiring new directions of thought, to take technology beyond what the Transformer architecture is capable of. Although people are impressed with GPT, the fact remains that many people don't really understand how it works. It is necessary for students to build AI from scratch and build tools to observe how "intelligence" emerges as the AI is trained with more data. Even if working with a smaller amount of training data, such efforts (when built with proper experimental tools) could help people understand the basics of how the pseudo-intelligence manifests itself, and how to improve it. This is crucial not just for designing better architecture, but also for designing better primitives. VL JEPA is one thought process in the direction of change, similar to my thesis proposal of the universal event model. Chronos 2 is another technique, which is very similar to my thesis proposal of partial contexts.
2. Fundamental changes in academia: The way academia encourages research, needs to be modified to help scholars truly explore and innovate/invent new technology. This requires better decisions on stipends, a focus on quality over quantity (paper publishing), and access to good computing power. It is crucial to connect researchers from various fields so that they can brainstorm and ideate, based on the various innovations each of them created in their respective fields of hardware, biological research, and software. It would help to incentivize innovation in more directions. Not all of it may be fruitful, but the exploration of ideas is necessary.
3. Corporations and governments creating leisure: While competition is good, the way it is currently being done is ending up creating burnouts instead of fostering healthy work environments. The manner in which the economy is designed itself may need a fundamental change, to enable countries to cooperate and take civilization forward.
Humans have achieved a civilizational leap in technology with the current advancement of AI. I hope the magnanimity of people among us would propel AI to a level that can help us create a heaven on Earth and beyond the solar system. To do that, we need to build a healthy foundation that is very different from the existing local optima that evolution is stuck at. Perhaps one way to progress would be to incorporate artificial intelligence and machines into the human body and create super-humans.
There were more opinions shared in part 2 of the hopes for AI.
For now, let us see what 2026 has to offer!
No comments:
Post a Comment