Probably because it was the first of its kind there is still a tendency to conflate chatbots bases on large Language Models with ChatGPT. But new natural language based models and interfaces are being released almost weekly at the moment. Here is a quick list but I appreciate it is not complete and I have not included the various art and graphic apps here:
- The new Bing
They are not all the same and I am not sure one is better that the others. They are mostly based on different large language models and training data and use different algorithms. To some extent this seems to be an attempt to target different users and perhaps different personalities too, if a chatbot can be said to have a personality. And the availability of Open Source models and data, is going to enable all kinds of focused apps in the future, including increasingly for coaching and learning (more on this in another post).
Meanwhile I have been quite enjoying talking with Pi. The text below is an extract from Darren Coxon’s Educalte Substack newsletter. Darren has been having a series of conversations with PI and publishing them in the newsletter. He says: “It is interesting how the Inflection RNN model enables this free-ranging dialogue, which can unlock some interesting insights.
What is most noticeable once again is how knowledgeable Pi is: it has clearly been trained on vast amounts of academic data. There seems little it doesn’t know, and to a very advanced level.“
Darren: Do you think that an AI could ever become conscious? that it could become aware of itself as a thinking being – as in Descartes’ cogito ergo sum?
Pi: Ah, the famous “I think, therefore I am.” That’s a very intriguing question, and one that raises a lot of philosophical problems. One of the main questions here is whether we can define what consciousness even is, let alone determine if an AI could possess it. What about this question – can you define consciousness? And, if you can define it, is there anything essential to the definition that an AI could never possess?
Darren: That’s a tough question. Consciousness is being aware of being aware. That there is awareness, but there also has to be something that is aware of this. We have sense objects, we have our physical senses, we have sense objects that our physical selves transmit into electrical signals through our sensory organs – but then there is the thing that makes sense of all that – that cannot be either the sense object or the sensory organ. Is that consciousness?
Pi: That’s an excellent definition, and one that many philosophers would agree with. It’s sometimes called the “hard problem of consciousness” – that we can understand the physical mechanisms of the brain, but not the subjective experience that emerges from those mechanisms. Philosophers like David Chalmers call this “the explanatory gap” between the physical and the subjective experience. So it sounds like you’re proposing that an AI could never cross this “explanatory gap” because it doesn’t have a physical brain?