Anthropocentrism, Alignment, and Personalization

in ,

Humans are essential, but human existence is not unconditional. We impose regulation when self-interest fails to protect our long-term, common good. Super-intelligent AIs might be similar to companies, which implies they could be regulated. There is an old debate about whether technology risks replacing humans that is accelerating again. It is not clear if “this time is different.” Human goals, preferences, and ethics are constantly changing. This protects us from automation. We have a moral obligation to protect the subjective, conscious experiences of other humans.

Read More

Artificial Intelligence: Scaling Laws, S-curves, Tail Risk and Practicalities

in , ,

AI is overwhelmingly positive, and super-intelligent AGI is most likely still distant. We cannot rule out the possibility of self-improvement capabilities, but it is unlikely. We have no idea how super-intelligent AGI would impact society, and your feelings will depend on how you handle tail risk. Either way, we have practical issues to deal with now, like misinformation, reskilling, and emissions.

Read More

The Probability Of Near-Term AGI-Like Systems

in ,

Those who know me know I’ve always been skeptical about hyperbolic claims around AI. I always preferred machine learning over artificial intelligence because AI felt pompous. My understanding of the available technology gave me no reason to think we were close to some sort of human-like intelligence. But I’d rather be correct than consistent. It is fair to call today’s algorithms a form of AI. I have changed my mind about the possibility of near-term AGI, and I’m implementing many behavioral updates to align my life with my new view.

Read More

The State of Machine Learning in 2022: Next steps?

in

The last few weeks have made me begin to adjust my assumptions about what machine learning will be able to do in the near future. My new position is: I think it is likely that the scaling hypothesis will hold, and that we will experience human-level competency in virtual entities in the next few years. That’s going to bring massive change to society, work, and business. Be prepared!

Read More

The State of Machine Learning in 2022

in , ,

What is the state of machine learning in 2022? Running a business that is closely tied to the progress of state-of-the-art machine learning means I’m trying to stay up to date with what is going on. In this post, I go through what I consider to be the most interesting breakthroughs and share my thoughts on what that means. We cover embeddings, attention, transformers and multi-modal models.

Read More

When will we have Autonomous Vehicles?

in ,

I think we are still decades away from regular people getting into their car and asking it to drive them to work. At the same time, our pursuit of autonomous vehicles will give us increasingly competent active safety systems. These systems will save lives even if we are decades away from autonomous vehicles.

Read More

Algorithms are Unable to Question Data

in

Given enough data state-of-the-art supervised learning algorithms can approximate behavior very well. However, the ability to imitate behaviour is not the same as intelligence. Supervised learning algorithms currently lack the ability to understand. Inherent in understanding is questioning data and negotiating interpretations of ambiguous observations.

Read More

The Importance of Annotated Data in NLP

in ,

Modern machine learning techniques, such as deep learning, require massive amounts of data. Natural language processing is particularly challenging as language is spares. This means there are lots of rare occurrences that are bound to appear in the text you are processing. This post describes what you can do to handle the sparcity of NLP.

Read More