# Constraints on AI and robotic advances

Go back to [[Week 4 - Introduction]] or see the [[Main AI Page]]

---

## Kevin Kelly on AI and Robots

The breakthrough that has not yet happened that will completely rearrange the current landscape of AI is using an extremely small dataset to train AI systems. Right now, AI requires very large training data sets to learn. And we have proof in the human toddler that we can actually have learning with very small data sets. Somebody in the future will figure out how to do that well. That will be a really huge shift, and it will be very liberating in many ways.

**You’ve said that at some point we'll all have personal robots. How far away do you think that is?**

Well, if the definition of a robot is an autonomous humanoid-shaped mobile robot with limbs, I think we're more than 30 years from that. And it’s not so much about the AI, which I think will be much sooner—it's that it’s an engineering challenge. A human brain runs on 100 watts and the human body is a quarter horsepower. And in terms of energy efficiency, we're not even within orders of magnitude near that efficiency. The kinds of AI we're running are incredibly inefficient compared to the brain’s energy.

What’s next for AI - Kevin Kelly. (2015, September 11). IBM Cognitive Advantage Reports. http://www.ibm.com/watson/advantage-reports/future-of-artificial-intelligence/kevin-kelly.html

---

## Yoshua Bengio on unsupervised vs supervised learning

Our main industrial systems derived from deep learning—such as speech recognition, machine translation, image search, self-driving cars, vision systems for blind people, etc.—take advantage of progress in deep supervised learning, yet humans are very good at unsupervised learning, and we need to make substantial progress in that direction to approach human-level AI

However, many hard problems remain open, like how to allow AI systems to automatically learn high-level semantics—in other words, how to represent more abstract concepts and the meaning of not just words but more complex thoughts, such as those we express in a phrase, a sentence or a paragraph.

## Yoshua Bengio on what needs to be unlocked before we can move ahead

**What will unlock the ability for computers to reason at the level of a human adult? How far off do you think this is?**

Difficult to say. Clearly one obstacle is simply to scale the current approaches. Another is to revisit the question of knowledge representation. Knowledge graphs were designed to be curated by humans, but we may need to learn other forms of representation which are more amenable to being extracted from data automatically—from documents, for example— by deep neural nets.

What’s next for AI - Yoshua Bengio. (2015, September 11). IBM Cognitive Advantage Reports. http://www.ibm.com/watson/advantage-reports/future-of-artificial-intelligence/yoshua-bengio.html

---

## Margeret Boden on the next questions to answer in the field of AI

Machines are very much better at understanding than they were before. They’re able to solve subtle things, pick up nuances in word usage, and so forth. But none of that is grounded in true human understanding. We need to understand what the brain is doing and how it’s doing it. Deep learning and other advances continue to generate useful applications, but considerable work remains at the analytical level to understand how human cognition works in supporting problem-solving and critical thinking and creativity.

**You don’t subscribe to the idea of the “singularity” then, the idea that computing will reach or match the level of human intelligence?**

The human mind is the most advanced system we have, and language in particular is hugely complex. AI and computing have progressed tremendously, but the capabilities remain quite limited. I doubt that computing will ever fully replicate the potential of the human mind, certainly not within the next few decades.

What’s next for AI - Margaret Boden. (2015, September 11). IBM Cognitive Advantage Reports. http://www.ibm.com/watson/advantage-reports/future-of-artificial-intelligence/margaret-boden.html