Interview with Gary Marcus about AI and ChatGPT

frankbell · 01-15-2023, 08:18 PM

On the January 1st edition of Le Show, Harry Shearer interviewed Gary Marcus about AI ChatGPT. I think you all might find it an interesting listen. The relevant portion starts at about the 20 minute mark.

sundialsvcs · 01-17-2023, 02:42 PM

But, you might also be very interested to read the recent postings in Bruce Schnierer's CryptoGRAM newsletter which apparently discuss enormous security holes in "ChatGPT," to the point that it might actually be a Trojan Horse.

frankbell · 01-17-2023, 08:15 PM

Thanks. I subscribe to his RSS feed.

Guess I'll have subscribe to his newsletter also!

I see that ChatGPT is getting a lot of hype. There was even a comic in my newspaper about it today.

sundialsvcs · 01-18-2023, 12:36 PM

You can read Bruce's "CryptoGRAMs" anytime you like on his website.

"AI" is getting a lot of hype these days – none of which seems to discuss the actual technology.

disk_bearing · 01-30-2023, 06:41 PM

I recommend using Chat GPT as a tool for learning things faster, and as a
preview of our frightening and fascinating future.
It's very fun!

via chat.openai.com

frankbell · 01-30-2023, 09:12 PM

You all might find this experiment with ChatGPT interesting.

It may be artificial, but, when put to the test, it doesn't seem all that intelligent.

boughtonp · 01-31-2023, 09:21 AM

Quote:

Originally Posted by Charles Babbage, Passages from the Life of a Philosopher (1864), ch. 5 "Difference Engine No. 1"

On two occasions I have been asked, — "Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?" In one case a member of the Upper, and in the other a member of the Lower, House put this question. I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.

I have a similar struggle to understand the people who get so enthused over ChatGPT.

Those that haven't listened to the Gary Marcus interview linked above might prefer to read this: https://garymarcus.substack.com/p/ais-jurassic-park-moment

teckk · 01-31-2023, 01:18 PM

Only thing I have ever got from chatgpt is something similar to this which I just tried.
https://0x0.st/oFGZ.png

Quote:

I recommend using Chat GPT as a tool for learning things faster,

One would have to be able to access it first. It's a nothing here so far.

teckk · 01-31-2023, 01:22 PM

Here is what I get

Code:

dig openai.com

; <<>> DiG 9.18.9 <<>> openai.com
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 21081
;; flags: qr rd ra; QUERY: 1, ANSWER: 2, AUTHORITY: 0, ADDITIONAL: 1

;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 512
;; QUESTION SECTION:
;openai.com.                    IN      A

;; ANSWER SECTION:
openai.com.             11      IN      A       13.107.237.51
openai.com.             11      IN      A       13.107.238.51

;; Query time: 10 msec
;; SERVER: 8.8.8.8#53(8.8.8.8) (UDP)
;; WHEN: Tue Jan 31 13:21:38 CST 2023
;; MSG SIZE  rcvd: 71

It's never "not busy". That's a Microsoft server.

teckk · 01-31-2023, 01:40 PM

I'm using my own web browsers on my linux machines. So, to make sure that it was not me, or a bug in webkit2gtk or QtWebengine, I fired up a windows10 notebook and tried to access it with Microsoft edge. Got the same busy, come back later page.

So this is a nothing sandwich so far here.

frankbell · 02-06-2023, 09:08 PM

Google is taking on ChatGPT: https://www.bloomberg.com/news/artic...-early-testers

frankbell · 02-07-2023, 08:08 PM

To belabor this topic, here's another interesting article I found: Matt Grawitch, a university professor, points out the misdirection play going on here. An brief excerpt:

Quote:

. . . coherence is not synonymous with accuracy.

ntubski · 02-07-2023, 09:30 PM

Quote:

Originally Posted by frankbell

To belabor this topic, here's another interesting article I found: Matt Grawitch, a university professor, points out the misdirection play going on here. An brief excerpt:

Hmm, this guy seems kind of clueless? He says "ChatGPT is more constrained than a human bullsh*tter because it cannot simply make stuff up (i.e., it must have data from which to draw to offer its responses)" and "ChatGPT is constrained to sharing information that it can curate from various sources to which it has access." which are both false: ChatGPT can definitely make stuff up (that is, output false facts), and it doesn't really have access to any sources (the training data affects the model's weights, which is sort of analagous to a human vaguely remembering reading things, but it is in no way direct access).

teckk · 02-08-2023, 10:19 AM

I can see where it would be useful for something like this:
https://m.youtube.com/watch?v=tFx5v8H6VXk

hish2021 · 02-09-2023, 08:03 PM

Here's a video on the presentations by Microsoft and Google: https://www.youtube.com/watch?v=5X1O5AS4nTc