News and What's New in AI

jwhjr17 said:

brianlux said:

PJ_Soul said:

brianlux said:

benjs said:

I've used ChatGPT to successfully write complex dialect-specific SQL queries with particular end goals in mind. I've also written business and/or software specs with it, drafted MVP designs for software, and am currently using it in conjunction with a development AI tool called Warp to write two different apps. 

I feel that the output is conservatively that of a $100,000 annual salary developer, and probably at least a 2x efficiency boost to a competent developer based on the time-savings.

Critical to understand:
1. The more explicitly and specifically you can outline objectives to these tools, the more likely they are to get it right
2. It relies on a sufficient corpus of data to yield good results, so these tools are unlikely to yield good results when working with a new coding tool/concept/language
3. It is not error-proof. If you don't have a (sometimes basic, sometimes more) understanding of programming logic, results will vary
4. It is not 100% optimal. Just as above, if you can't recognize suboptimal code or propose better ways, you may end up with suboptimal app performance
5. It will not give unbiased opinions (even based on evidence). It is more likely to explain to you why you're right about something, than why you're wrong about something
6. Tools are getting better at addressing the first five points rapidly, and a developer who doesn't use these tools is already disadvantaged compared to a developer who does

I'm just beginning to try to understand AI, Ben, not as a fan nor as a detractor, but simply because it is here to stay and until the power runs out, it will eventually influence basically everything. So my question is this: If we were to define "true AI" as something that, "Exhibits a broad spectrum of cognitive abilities, comparable to human intelligence, allowing it to understand, and to reason on its own, and learn across diverse domains", is ChatGPT true AI?

Yeah, ChatGTP is true AI. It does a lot more than just answer questions. For example, I could just enter a job posting for a job I'm applying for and copy and paste my resume, and ChatGPT will create a perfect unique cover letter for me suitable only for that job, created to try and get me that interview. And that's just a really basic use for it. 

I'm fairly certain what you are talking about is not what I understand true AI to be.  If you go to the video I posted above with the conversation between Sir Stephen Fry and Yuval Hahari (well worth the half hour to watch in any case), Fry and Harari talk about how true AI can do things we cannot.  At around 6:02 Fry says, "...if it could only match what we did, it would not be a tool..." and how AI can outmatch what we can do.  ChatGPT may be able to write your resume faster than you, but you are intelligent enough to  be able to write a resume as well as ChatGPT.  

I think we are probably on the brink of, or even at, the point of true AI, but my impression is that the term "AI" gets tossed around a bit too easily.  The things we describe as AI are things that the machine does on our command.  True AI is or will be something capable of doing things without commands and (again, this is in the video) when true AI is fully functional, we will be living in a hybrid society.  The ramifications of that are fascinating at best, disturbing or even disastrous for humans at worse.  

I'm not trying to be fear mongering here, but I think if you watch the Fry/Harari video as well as this shorter one:

https://www.ted.com/talks/tristan_harris_why_ai_is_our_ultimate_test_and_greatest_invitation

you will understand that I am not (as jwhjr17 above seems to be implying) simply dear mongering.  I'm just observing what we are creating.  

@brianlux I apologize if you thought I was implying that you were fear mongering as that was not my intent. 

Honestly, I don't remember.  Oh the joys of aging. 

And I apologize if I misunderstood.  All's good.  Thanks for the message!