I really hope ChatGPT or another, similarly capable model will continue to be available for free in quantity sufficient for personal usage (copious amounts of it). I almost don't use my free-tier GPT-3 because of a psychological barrier of being aware there is a limit, even if it should be high enough.<p>ChatGPT has been on my mind ever since it was made available for all kinds of possible questions from philosophy, nature, universe, law, linguistics and translation, writing code, diagrams, graphs, debugging code and systems. I've learned to intuit what kind of question will provide an interesting answer. I'm still sometimes struggling to reformulate questions it should be able to answer but provides canned responses or made up answers.<p>I appreciate it for creative suggestions (think "gifts for ...", "questions for ...") and for summarizing/explaining content interactively. I admire its ability to appear to be reasoning without actually reasoning.<p>I think the next steps for development would be:<p><pre><code> * Pairing it with some logical reasoning and computing engine so that it doesn't make elementary mistakes (Wolfram Alpha comes to mind as state of the art).
* Pairing it with a knowledge graph like Wikidata.
* Letting it browse the web to get up-to-date information in real time and process it.
* Adding more visualization modes (it generates tables when asked) like charts and graphs (mental maps could be a popular usage), perhaps interactive ones. You can currently do this surprisingly successfully by generating the text, then plugging it to a visualization tool. There could be an automatic feedback loop where the tool points out errors and the model fixes them so that the user doesn't have to ask explicitly.</code></pre>