Three years later, ChatGPT is still not what it should be – and probably never will be

Three years ago, on November 30, 2022, ChatGPT was released. It was one of the fastest growing consumer products in history and got more press than God. But I think it’s fair to argue that it’s not what is often portrayed and probably never will be.

Before I delve deeper, I want to be absolutely clear about four of my core beliefs that are often misrepresented:

1. I believe that artificial general intelligence (AGI) is achievable.

2. I believe there is at least a chance that artificial general intelligence will have a large net benefit to society.

3. I just don’t think large language models like ChatGPT will get us there. (While I believe they have their uses, I worry about their costs to society, such as bias, cybersecurity, misinformation, non-consensual deepfake porn, copyright theft, energy and water consumption, the gradual de-shitification of the Internet, the severe impact on higher education, and so on.)

4. I think that the recurring core technical problems we have…