Coding, Data Science, A.I., Robots |

  • Thread starter Thread starter nycfan
  • Start date Start date
  • Replies: 479
  • Views: 15K
  • Off-Topic 
NCK9 said:
Is chatGPT smart enough to put and salary < 100000 into the SQL query or does it do it in the code while it iterates over the results?
Click to expand...
Good question! Scary if so. I certainly didn't look, didn't even think of that... I don't have a mind for efficient architecture or algorithms.

There's some quote about QA resources needing to be N times your programming resources and that number is obviously going to need to skyrocket for anything serious.
 


🎁 —> https://www.wsj.com/politics/policy...d?st=W5SppR&reflink=desktopwebshare_permalink

“… The potential executive order comes as administration officials try to impose a moratorium on state AI laws or policy pre-empting state rules that is part of the federal defense spending bill working its way through Congress.

A similar effort earlier in the year failed after former Trump adviser Steve Bannon and some senators including Sen. Marsha Blackburn (R., Tenn.) opposed it saying it undermined workers and the safety of children online.

… State laws in New York, California and elsewhere have irked administration officials like White House AI Czar David Sacks and business executives who have called for federal action. They say a patchwork of state rules makes it impossible to comply and slows down innovation. …”
 
gregh1 said:
Good question! Scary if so. I certainly didn't look, didn't even think of that... I don't have a mind for efficient architecture or algorithms.

There's some quote about QA resources needing to be N times your programming resources and that number is obviously going to need to skyrocket for anything serious.
Click to expand...
It is absolutely "capable" of creating the WHERE clause. But it is not remotely "smart" in any sense we understand the word. Anything for which example code abounds on the internet that can be scraped up into the training data can be detected as a pattern that can be replicated in the form of output by the model. Because this works by tokenization, variables like 100000 are handled very well (what is a variable if not a token?).

Long story short LLMs can spit out customized version of anything they've seen multiple examples of in the training data. That's the best way to think about it.
 
Who would ever choose Grok among the available AI choices? Musk has been fairly transparent about the built in biases.
 
farce© said:
Who would ever choose Grok among the available AI choices? Musk has been fairly transparent about the built in biases.
Click to expand...
It's not my favorite LLM. It's my second favorite open source LLM (the older versions) not including the Chinese models, which are really good but uncertain on their security.

I do think it's the best at free image and Video generation. Sora from Chat GPT is the best but they are really expensive. Veo from google is also really good and much more reasonably priced but I don't think it's quite as good as Grok. But google just released a new one that I haven't played around with.

I believe Grok might also be the only open source that lets you generate porn.

But the leaderboard for these models changes every month or two so somebody that is on top today won't be on top in 6 to 12 months if they don't keep innovating.
 
With this coming out, Trump should do a rally or something to show appreciation for all the newly discovered support he has in Nigeria.
 
You must log in or register to reply here.
Back
Top