Is GPT-3 conscious? The debate about AI gets more intense

GPT-3
Image credit | studiostoks/bigstockphoto.com

When an AI machine issues a rebuttal about its consciousness, you know you are living in interesting times. The machine, called GPT-3 (Generative Pre-Trained Transformer, version 3) is certainly a step or two closer.

GPT-3 has now ‘consumed’ 45 terabytes of data. To put that into context, the whole of Wikipedia accounts for less than half a percent of the information that GPT-3 has trained with to make its decisions.

The ‘awareness’ issue around GPT-3 arose when one Professor Chalmers said that if a worm with 306 neurons is conscious then he would be open to the idea that a ‘machine’ with 175 billion parameters is also conscious.

Whether or not you believe that GPT-3 is indeed conscious (and the computer itself says ‘no’) you have to wonder how far AI can go. Although AI has been around for a long time, it is now evolving fast as we discover new and faster ways of training it.

But GPT-3 is clear about what it is.

“I am not self-aware. I am not conscious. I can’t feel pain. I don’t enjoy anything,” it said in response to the statement by Chalmers.

Then it went on to say “the only reason I am responding is to defend my honour.”

Which seems to suggest that it has a sense of humour (and honour). This might have been learnt from the 45 terabytes that it can draw on, or it could mean that in fact, it is conscious and therefore conscious of the fact that it has a sense of humour.

This slightly flippant issue does pose the question: will we get to Artificial General Intelligence (AGI), where AI evolves from a linear ‘existence’ to one where it can reason in the same way that we can.

MIT researchers are sceptical that we can ever get to AGI, yet San Francisco based OpenAI, who built GPT-3, believe we can get there and it will open the gates to a ‘very powerful future.’

The problem with AI lies not with the machine but those who control the machine (and the bias they bring) and there are many people out there comparing AI to the evolution of the atomic bomb. Splitting an atom was a major scientific breakthrough, using it to kill a lot of people was not.

Whatever the future has in store for AI and GPT-3, there is no doubt that it will be a powerful tool and much good will come from its implementation – in business, education, health and life. There is also no doubt that it can be used for surveillance, hacking and highly efficient warfare.

We have a choice.

1 Comment

  1. “We have a choice.” There is no “we” making the choice. The person who wields a tool that can be used destructively or productively is the one who decides how to use it. Similarly, “we” don’t decide whether to use a gun in self-defense or to commit murder.

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.