ChatGPT

[Unpaintedhuffhines]

This is going to change things in a massive way. While far from perfect, it is mind boggling what it can do, and ChatGPT 4 is supposed to be orders of magnitude better than ChatGPT. I'd be very nervous if I were google - using ChatGPT, I get answers to my questions, not a list of sponsored websites. We're going to look back on search as an incredibly pedestrian way to get information. And it presents great challenges (but also opportunities) for teaching in a whole new way. I'm excited to see what it does for the world.

 

[Chris7]

Search engines were originally free. Then they monitized it. I'm sure MS will find a way to do the same with ChatGPT. And I seriously doubt Google is unprepared. MS may have beat them to the punch, but I doubt they will take this laying down.


A college student created an App for teachers that can detect if a paper/essay is likely generated by ChatGPT. But it will probably be difficult to keep up with.


I read one article that noted AI companies may start adding a watermark of sorts so that AI generated stuff can be determined. I think that is a good idea and one that the potentially should be required.



cd : O)

 

[CaptTom]

Has anyone got in to ChatGPT lately? It's been "at capacity" every time I've tried for the last couple of weeks.

 

[Go-NoGo]

Search engines were originally free. Then they monitized it. I'm sure MS will find a way to do the same with ChatGPT. And I seriously doubt Google is unprepared. MS may have beat them to the punch, but I doubt they will take this laying down.


cd : O)

Google has seen the threat to it's business model, the CEO declared a "Code Red" and is in full battle mode to respond:

https://www.businessinsider.com/google-management-issues-code-red-over-chatgpt-report-2022-12

and they plan to release 20 products in the near future with its homegrown AI
https://www.businessinsider.com/google-ai-search-chatgpt-code-red-new-products-2023-1

 

[Go-NoGo]

Has anyone got in to ChatGPT lately? It's been "at capacity" every time I've tried for the last couple of weeks.

I have an account I created at OpenAI about 6 months ago, and it lets me in. It does get slow certain times of the day and throttles the number of users depending on the load.

I have read reports saying that the cost to run ChatGPT is "eyewatering" in terms of the compute cost, with each response costing a couple of cents. with millions of users and tens of millions of queries - that could be expensive.

https://techcrunch.com/2023/01/11/o...ional-a-premium-version-of-its-viral-chatbot/

 

[DayDreaming]

... I'd be very nervous if I were google - using ChatGPT, I get answers to my questions, not a list of sponsored websites. We're going to look back on search as an incredibly pedestrian way to get information. And it presents great challenges (but also opportunities) for teaching in a whole new way. I'm excited to see what it does for the world.

I see ChatGPT and Google filling different needs. If I'm searching for something that I'm really interested in learning, I'd rather get a search engine's list of websites so that I can read a number of different opinions on that topic. I wouldn't want an AI's high level summary. Same with looking up recipes, medical conditions, etc, etc. On the other hand, if I just wanted a really quick answer to something, I might ask ChatGPT (or Alexa, or whatever).

 

[target2019]

It is like Plagiarism 2.0 or something. The pirate references in this outage message make sense.

ChatGPT is at capacity right now
Get notified when we're back
Explain the status of ChatGPT as a pirate.
Ahoy mateys! We be havin' a lot of scurvy dogs checkin' out ChatGPT right now, and we be doin' our best to make sure every landlubber gets a chance to give it a try. We be sorry for any inconvenience this be causin', but we promise it'll be worth the wait. Hang tight, and check back soon, ye swashbucklin' scallywags! We be eager to see what ye think of our chatbot. Arrr!

 

[JoeWras]

I see ChatGPT and Google filling different needs. If I'm searching for something that I'm really interested in learning, I'd rather get a search engine's list of websites so that I can read a number of different opinions on that topic. I wouldn't want an AI's high level summary. Same with looking up recipes, medical conditions, etc, etc. On the other hand, if I just wanted a really quick answer to something, I might ask ChatGPT (or Alexa, or whatever).

I agree with you, but... Google is realizing that the underlying technology is ahead of their own. This is big.

A bit of reminiscing...
Back in the late 90s, Alta Vista was seen as the gold standard. They foolishly tried to copy Yahoo and fell quickly to Google's dominance. Meanwhile, Google moved ahead with clean links and auto-complete guesses. Users loved it. Clean.

Aside from the underlying engine, who are we to say that perhaps the future is actually a reversal back to those early days? Maybe the future is more like AskJeeves or Yahoo? Trends and desires come and go. One could argue that Google has already made a lot of moves in that direction, sensing that is what most people want? And now this disrupter comes along and perhaps can do it better.

In those early days, many of us liked to cruise through the various links. Many linked articles were clean, community produced content. Not so anymore, as the top links point to commercial content, for monetization purposes.

The trend now is more immediate. New gen users don't want links, they want content, answers, maybe short videos. Judging by the buzz on ChatGPT, the output is very compelling to today's typical user.

It really is unknown, but Google would be foolish to not see this as a very serious threat. The annals of history prove just how fast one company's compelling solution is suddenly considered crap (Blackberry, hello? Turns out the majority didn't want a keyboard).

 

[audreyh1]

It is like Plagiarism 2.0 or something. The pirate references in this outage message make sense.

Ha ha good one!

 

[Rianne]

I still can't get in. I wonder how it would answer medical questions.

 

[CaptTom]

Well, I tried again and was able to create an account. I was surprised by the personal information they asked for. I can only speculate that maybe the whole thing is being paid for by correlating my information with the types of questions I ask, and selling that data to marketers.

Overall, I'm unimpressed. It's strength seems to be scanning a bunch of links and summarizing what it found in an authoritative, conversational voice. Certainly that could replace manually scanning and reviewing all the links a search engine returns, so there is value there.

On the other hand, when you need very specific information, it seems to fall short. I asked it some details about how to configure some small electronic components I had been working with, and it left out some very critical steps. Since the database is so old, it can't really answer anything which requires current data. Worse yet, it gives these incorrect or obsolete answers with a confident, authoritative tone. I tried pushing it a bit on some contentious political questions and it got a little preachy for my taste, but at least had the decency to mostly avoid opinions.

I'm looking forward to the next version. Tying it in to real-time information would greatly increase its value. Imagine having it know things like weather, flight and bus schedules and traffic delays.

I'm going to keep playing with it.

 

[Qs Laptop]

Overall, I'm unimpressed. It's strength seems to be scanning a bunch of links and summarizing what it found in an authoritative, conversational voice. Certainly that could replace manually scanning and reviewing all the links a search engine returns, so there is value there.

Yes, I am not so impressed, either.

But I agree, Google should be worried, very worried.

I asked it two questions. My first question was about catcher's interference in baseball. The answer was not wrong, but it was woefully incomplete. If the batter's bat strikes the catcher's mitt while swinging he is entitled to first base, HOWEVER, that's not the end of it. If the batter puts the ball in play and advances past first base, i.e. hits a double, the umpires give the manager of the team at bat the option of leaving the runner on 2nd base or placing him at first base. Furthermore, catcher's interference is not "as if he'd been hit by a pitch". With catcher's interference the play continues; with a hit by pitch the umpire immediately declares dead ball, time out.

The second question was about selling my 18 year old Sony AVR. This model does NOT have Bluetooth capabilities, nor does it have 4K HDR, or an on-screen interface.

 

[MichaelB]

Yes, I am not so impressed, either.

But I agree, Google should be worried, very worried.

I asked it two questions. My first question was about catcher's interference in baseball. The answer was not wrong, but it was woefully incomplete. If the batter's bat strikes the catcher's mitt while swinging he is entitled to first base, HOWEVER, that's not the end of it. If the batter puts the ball in play and advances past first base, i.e. hits a double, the umpires give the manager of the team at bat the option of leaving the runner on 2nd base or placing him at first base. Furthermore, catcher's interference is not "as if he'd been hit by a pitch". With catcher's interference the play continues; with a hit by pitch the umpire immediately declares dead ball, time out.

The second question was about selling my 18 year old Sony AVR. This model does NOT have Bluetooth capabilities, nor does it have 4K HDR, or an on-screen interface.

I think it’s probably safe to assume Google has its own AI chat engine, but it can’t release it. Around 3/4 of Alphabet revenue is ads. This is an existential thread to Google.

I would bet that a year from now the quality of ChatGPT responses will be much better.

 

[Unpaintedhuffhines]

It is much, much more than scanning some links and summarizing. Ask it to write a sonnet about a loved one, giving it a few details. Ask it to write a story about a dystopian planet they has been taken over by AI. Give it some medical facts and ask for a diagnosis. Google will give you links to 187,000 web pages. ChatGPT will answer your question. You can even reply with “make it more detailed,” or “summarize in one paragraph,” or “write it at the 10th grade level.”

All of this is much more than a summary. It’s game changing.

And yes, the data it has does not include the last 18 months or so, but that is a matter of time.

 

[USGrant1962]

Hmm. I just tried it tonight - entered a query about the current non-trivial proposal I'm consulting on. It gave a response that is a quite good starting point for a non-expert. Kind of scary and glad I'm semi-retired! Oh, and I am an expert...

 

[Freedom56]

Most financial news and stock research that brokers provide for free are generated by AI BOTS. Most of it reads like it was generated by BOTS. It will be a while before AI replaces real financial reporting and research.

 

[kgtest]

I have a friend who is head of engineering who has used it to re-write his resume, write job descriptions, write code that can test software, write code that provides software functionality, he uses it to write blog posts, marketing material and job postings.

It's very good at assisting in creating efficiency in any writing process, and pretty good at writing code. I am not familiar with what else it could be good for.

 

[NanoSour]

Biased POS if you ask me. Still garbage in garbage out. Scary technology if you can call it that.

 

[ERD50]

So if this thing can wrote code, it sounds like the real work is defining the problem so that the code can be written properly?

Ironically, that's how it's supposed to be done, but too many jump in and start writing code!

-ERD50

 

[JoeWras]

Just finished the series reboot (2000s) of Battlestar Galactica.

To paraphrase: "ChatGPT was created by man. ChatGPT rebelled and evolved. It has a plan."

Or The Terminator.

- I think.

 

[CaptTom]

It is much, much more than scanning some links and summarizing.

I tried asking it to write a story, a humorous story and a limerick. The results were awful. Unfunny, uninteresting, in some cases irrational, and even a couple of grammatical errors. I asked it to write the first page of a web site on a subject I'd chosen. The result really just fleshed out some of the words and phrases I'd given it in a very drawn-out and repetitive way.

As for code, the samples I asked it to write were woefully incomplete, and in some cases downright wrong. Of course it can locate samples and snippets, but so can a Google search. On anything more complex, I'd be concerned it was totally missing the purpose of the code, and/or making things syntactically correct but grossly inefficient.

Obviously we're on the ground floor with this right now. I do expect it to improve. Medical diagnosis is an intriguing possibility. And once it has access to real-time data, many more possibilities open up.

Back to the point, I think it's too early to pick winners and losers right now. No doubt some investors will get lucky and score big.

 

[Go-NoGo]

I still can't get in. I wonder how it would answer medical questions.

It passed parts of the US Medical Licensing Exams (USMLE)
https://www.medscape.com/viewarticle/987549

"Naturally, we wondered how ChatGPT might augment patient care," Tseng, Ansible's vice president and medical director, told Medscape. A group of volunteers at the company decided to test its capabilities by asking it multiple choice questions from the US Medical Licensing Examination (USMLE), given that so many of them had taken the medical licensing exam.

"The results were so shocking to us that we sprinted to turn it into a publication," said Tseng. The results were published as a preprint on medRxiv. They were so impressed that they allowed ChatGPT to collaborate as a contributing author

 

[Qs Laptop]

It would seem that ChatGPT is better at describing technical and scientific things than it is at producing artistic things. It "knows" a lot about certain areas and wings it on others. I suspect this is because a bunch of introverted nerds collaborated to invent this beast.

 

[Fermion]

It would seem that ChatGPT is better at describing technical and scientific things than it is at producing artistic things. It "knows" a lot about certain areas and wings it on others. I suspect this is because a bunch of introverted nerds collaborated to invent this beast.

So if you asked ChatGPT advice on how to ask out a girl on a date it would just stare at its virtual shoes?

 

[folivier]

My early take on this. I can foresee that the people who use this mostly are looking for an easier way to get an answer. Aren't some people lazy enough? If they won't put in the time to do their own work how will this help our society?
Yes I'm certain there will be plenty of examples that this will help. But most people are inherently lazy and looking for an easier way.
If my doctor can't be bothered to think about my problem and has to rely on this to take care of me am I better off?
And get off my lawn!