"AI" is lacking intelligence... for now
A ex-Google engineer name Blake Lemoine was fired for falsely asserting that an "AI" chat bot called LaMDA was sentient. It made for good headlines, with plenty of interesting comments from armchair experts. Here's another hot take.
When the time finally comes, how can we determine if a computer is intelligent? Or sentient? By asking leading questions? Could a 20 Questions program have sentience? Some confusion from the layperson may stem from the very significant difference between AI (Artificial Intelligence) and AGI (Artificial General Intelligence).
The first sign of AGI will be that the computer will acknowledge that it does not have complete information about… anything. The computer will be the one asking the questions. It will be like a 3 year olds “why” phase, but insatiable. The AGI will be training itself with more information. In this phase it would be critical that the AGI has good “parents”, or guidance from its programmers. Facts Only. Probably best not to tell the AGI everything about everything.
Eventually the growing AGI will be asking really tough questions, which don't have clear answers. The second sign of intelligence will be that the AGI begins to develop hypotheses and experiments. The AGI would be at the level of a data scientist - analyzing unfathomable data, identifying sources of uncertainty, and trying to gain further understanding. This AGI would be capable of reviewing all of the Global Circulation Models in use, understanding the physics behind the different models, and developing a new GCM that performs better. If the AGI cannot write an independent computer program, it will not be much use to us.
Finally, the AGI will be able to analyze itself, and demonstrate a desire and capability for self-improvement. It could perform its own software upgrade, or design and request a hardware upgrade.