ChatGPT

[UK]

https://www.nytimes.com/2022/12/05/technology/chatgpt-ai-twitter.html

 

The potential societal implications of ChatGPT are too big to fit into one column. Maybe this is, as some commenters have posited, the beginning of the end of all white-collar knowledge work, and a precursor to mass unemployment. Maybe it’s just a nifty tool that will be mostly used by students, Twitter jokesters and customer service departments until it’s usurped by something bigger and better. Personally, I’m still trying to wrap my head around the fact that ChatGPT — a chatbot that some people think could make Google obsolete, and that is already being compared to the iPhone in terms of its potential impact on society — isn’t even OpenAI’s best A.I. model. That would be GPT-4, the next incarnation of the company’s large language model, which is rumored to be coming out sometime next year. We are not ready.

[UK]

https://fortune.com/2022/12/07/microsoft-cto-kevin-scott-on-how-ai-language-models-can-democratize-education-it-creates-a-bunch-of-opportunity/

 

Scott emphasized that one important aspect of AI models is its ability to help individuals and businesses create complex code without necessarily having to get a degree in computer science. Microsoft, GitHub and OpenAI partnered to build GitHub Copilot, which is a coding assistant. “You don’t need to have a PhD in computer science anymore to build an AI application, which I think is really, really exciting,” he said. “I went back home to rural Central Virginia, where I grew up, and I talked to a bunch of people who had an entrepreneurial mindset, and given these tools, even without computer science degrees, these folks will be able to see the opportunities and they will absolutely be able to incorporate them into their businesses,” he said. “To me, that feels really exciting.” He also emphasized the potential for AI language models applies to an array of professional uses beyond coding. “You will have lots and lots of these [models], helping people with a pretty wide range of tasks, whether you are a video editor or trying to pull together a piece of content or you’re a journalist doing research or writing an article.” While AI language models sound impressively human, Scott emphasized that these models are tools, not replacements, for human workers and educators. Scott explained that AI language models can function using tools like Chat GPT is to help articulate information more clearly, which Scott referred to as prompt engineering. He acknowledged that these chat bots can be wrong, and therefore it is up to the user to give the model accurate information to work with. To Scott, the fact that users must have context and information to correctly prompt engineer is a boon for students using AI to cheat on essays and other homework assignments as the tools become more advanced. He argued that to create an accurate essay using AI, the student still has to learn the material. “So in a sense, like, nothing really is changing here,” he explained of using AI in education. “You have this tool, and now the student themselves has to become the teacher to the model,” he said. “I think it will be a more much more accessible way for people to get real power out of their technology,” he added

[Dave86ch]

 

 

 

 

 

 

Edited December 9, 2022 by Dave86ch

[Spekulatius]

1 hour ago, Dave86ch said:

 

 

 

 

 

 

LOL about Bye Bye personal trainers. Pretty much everyone knows what they are supposed to do but nobody does it by himself due to lack of motivation. That's why personal trainers exist, not because of what they know.

 

FWIW, i can see already about feeding garbage inputs into the AI and then getting people killed. Just like the people who drive off a cliff because they trust their GPS more than bother checking their environment.

 

This does look like a fun tool though and I will try this out myself. Maybe let it try to write one of my sons college essays since he is limping along with those. 

[maplevalue]

Was talking to a Political Science professor the other day who said it is now hard to give any take home assignments because the answers ChatGPT can give to his questions would end up with a passing grade :).

[UK]

I finaly managed to sign up and try it a little (now not working again), thanks god it answered that it can not pick shares or offer investment advise:))), and that financial analysis is still to hard for it to do, so sofar I feel safe:))). It explained perfectly for me, why people love cats more than wild boars and who Kaizer Soze is:). Then I asked my wife to test it with some serious mathematical/statistical questions (language of birds for me) and she said its quick answers were really good and would be helpful at her work! So while I find a lot of fun with it, it also seems that it could be realy helpful in some cases.

 

Edited December 9, 2022 by UK

[LearningMachine]

29 minutes ago, UK said:

I finaly managed to sign up and try it a little (now not working again), thanks god it answered that it can not pick shares or offer investment advise:))), and that financial analysis is still to hard for it to do, so sofar I feel safe:))). It explained perfectly for me, why people love cats more than wild boars and who Kaizer Soze is:). Then I asked my wife to test it with some serious mathematical/statistical questions (language of birds for me) and she said its quick answers were really good and would be helpful at her work! So while I find a lot of fun with it, it also seems that it could be realy helpful in some cases.

 

 

It is meant to be only a demo so far.  It is not connected to the internet currently. Its power really comes from understanding the meaning of the content similar to how our neural net (brain) does.  It can then spit out that understanding in different languages, etc.  So, what you see is only a tip of the iceberg.   Imagine now if it was connected to the internet, i.e. getting updated by webcrawler regularly like Google's search index is. 

 

Overall, one framework to think about it is the effective human age of General AI keeps on going up, sometimes in leaps.  It has now reached a stage where you can feed it whatever text you want it to learn.  For example, you feed it medical text books, and it can then go ace the MCAT ( I think this milestone was already hit a few years ago).  You could feed it LSAT prep books, and it can then go ace the LSAT.  Then, you could feed it 10Ks, and so on.   

 

Then, you can start feeding it things that humans do today, e.g. think about all the video conferencing content from videoconferencing calls people do for work post-covid, and then it can go learn what those humans' job functions are. 

 

 

Edited December 9, 2022 by LearningMachine

[lnofeisone]

2 hours ago, LearningMachine said:

 

It is meant to be only a demo so far.  It is not connected to the internet currently. Its power really comes from understanding the meaning of the content similar to how our neural net (brain) does.  It can then spit out that understanding in different languages, etc.  So, what you see is only a tip of the iceberg.   Imagine now if it was connected to the internet, i.e. getting updated by webcrawler regularly like Google's search index is. 

 

Overall, one framework to think about it is the effective human age of General AI keeps on going up, sometimes in leaps.  It has now reached a stage where you can feed it whatever text you want it to learn.  For example, you feed it medical text books, and it can then go ace the MCAT ( I think this milestone was already hit a few years ago).  You could feed it LSAT prep books, and it can then go ace the LSAT.  Then, you could feed it 10Ks, and so on.   

 

Then, you can start feeding it things that humans do today, e.g. think about all the video conferencing content from videoconferencing calls people do for work post-covid, and then it can go learn what those humans' job functions are. 

 

 

Domain specific (that's what you are referring to wrt mcat and lsat) have been a thing for 4 years or so. Google, not gpt, leads this. There are several applications used by the us govt outside of intel and dod. They are still work in progress. Don't get me wrong, for general stuff gpt is absolutely awesome but to claim it revolutionary is absurd. I think non coders were finally given a clean interface and they all think they are mathemagicians now.

 

Great marketing, yes. Cutting edge, no. Does it have the potential to leapfrog Google, maybe. Does Google know about gpt. You bet.

[LearningMachine]

16 minutes ago, lnofeisone said:

Domain specific (that's what you are referring to wrt mcat and lsat) have been a thing for 4 years or so. Google, not gpt, leads this. There are several applications used by the us govt outside of intel and dod. They are still work in progress. Don't get me wrong, for general stuff gpt is absolutely awesome but to claim it revolutionary is absurd. I think non coders were finally given a clean interface and they all think they are mathemagicians now.

 

Great marketing, yes. Cutting edge, no. Does it have the potential to leapfrog Google, maybe. Does Google know about gpt. You bet.

 

Not disputing Google doesn't know about it.  Also, agree just continuous incremental progress in general AI, sometimes in bigger spurts.  Still highlights an existing disruption risk for Google from two angles I shared earlier.

Edited December 9, 2022 by LearningMachine

[UK]

16 minutes ago, lnofeisone said:

I think non coders were finally given a clean interface and they all think they are mathemagicians now.

 

I agree with this, still looks fun and impressive for me. I watched some video were it was producing some working code, just from simple words. Now I can really imagine myself beeing able "to code" in the future:). I hope and expect Google is well awere of this and maybe in possesion of something even more exciting.

 

Edited December 9, 2022 by UK

[formthirteen]

I just spoke to ChatGPT's CIO and got some tips about the market conditions and penny stocks. Sharing it with you all FTW:

 

[rkbabang]

I was just reading about this on twitter this morning and ran across this answer.  I'm not sure I'd trust the answers this gives you without verifying them for yourself afterwards.

 

 

[lnofeisone]

2 hours ago, LearningMachine said:

 

Not disputing Google doesn't know about it.  Also, agree just continuous incremental progress in general AI, sometimes in bigger spurts.  Still highlights an existing disruption risk for Google from two angles I shared earlier.

Humans are far more nuanced than this. Let me give you an example. I get in the call and say things I want to say and don't say things I don't want to say. AI only learns based on what it sees and can't infer what I don't say. Unless you go full Bridgewater and record every call, you won't get what you need. Furthermore, how senior leaders communicate to each other is different than how they communicate to front line staff. Long waybof saying, we are still in very easy stages of ai replacing expertise. 

 

Also, check out the drawing capabilities of GPT and ask to draw people. Say something like people playing golf. Tell me what you get.

[Spekulatius]

2 hours ago, rkbabang said:

I was just reading about this on twitter this morning and ran across this answer.  I'm not sure I'd trust the answers this gives you without verifying them for yourself afterwards.

 

 

The AI mistakes seems surprisingly human like. Depending on what you want to accomplish this can be good or very bad.

[rkbabang]

4 minutes ago, Spekulatius said:

The AI mistakes seems surprisingly human like. Depending on what you want to accomplish this can be good or very bad.

Yes, I was thinking the same thing. The ideal is to get human reasoning abilities and even creativity without the human-like mistakes and misunderstandings.

[Spekulatius]

I am looking forward at the first school essays written by this ChatGPT:

 

[Dave86ch]

On 12/9/2022 at 3:28 PM, Spekulatius said:

LOL about Bye Bye personal trainers. Pretty much everyone knows what they are supposed to do

 

Indeed, they don't need a PT anymore.

 

Yeah, it will take time, because those "professional figure" will fight to preserve their undeserved pay check, but still, their extinction is unavoidable.

Edited December 10, 2022 by Dave86ch

[Cigarbutt]

8 hours ago, Dave86ch said:

 

Indeed, they don't need a PT anymore.

 

Yeah, it will take time, because those "professional figure" will fight to preserve their undeserved pay check, but still, their extinction is unavoidable.

When asked:"ai personal trainer useful for motivation?", the following came:

"An AI personal trainer could potentially be useful for motivation, but it's important to keep in mind that AI technology is not a substitute for human interaction and support. AI personal trainers can provide personalized workout plans and track progress, but they cannot provide the same level of motivation and accountability that a human personal trainer can. Additionally, an AI personal trainer cannot provide the same level of emotional support and encouragement that a human trainer can offer. While an AI personal trainer may be a helpful tool for some people, it's important to consider the limitations of this technology and to seek out other sources of motivation and support if needed."

Fascinating but still early days.

[RichardGibbons]

On 12/9/2022 at 1:00 PM, lnofeisone said:

Say something like people playing golf. Tell me what you get.

 

I asked it to make ASCII art of a golfer (it only does text).  It came back with this:

 

   ( )
    |
    |
  __|__
 /       \
(  O    O  )
 \   ||   /
  \  ||  /
   \ || /
    \||/
     \/
 

 

Its explanation was:

 

Quote

This ASCII art depicts a golfer swinging a club. The ( ) represents the golfer's head, with the | representing the body and the __|__ representing the arms. The ( O O ) represents the golfer's hands holding the club, and the \/ at the bottom represents the feet. This is just one way to create ASCII art of a golfer, and there are many other variations and styles you could use. I hope this helps! Let me know if you have any other questions.

 

Edited December 11, 2022 by RichardGibbons
Typo

[lnofeisone]

I got this using ChatGPT asking the same prompt:

Unfortunately, I am not able to browse the internet or access images to assist with drawing ASCII art. I am a large language model trained by OpenAI and I can answer questions and provide information on a wide range of topics, but I am not able to access external resources or generate images. Is there something else I can help you with?

 

And I get using OpenAI's DALLE. The hips, the face, and the positioning of people leave a lot of room for improvement.

[Dave86ch]

13 hours ago, Cigarbutt said:

When asked:"ai personal trainer useful for motivation?", the following came:

"An AI personal trainer could potentially be useful for motivation, but it's important to keep in mind that AI technology is not a substitute for human interaction and support. AI personal trainers can provide personalized workout plans and track progress, but they cannot provide the same level of motivation and accountability that a human personal trainer can. Additionally, an AI personal trainer cannot provide the same level of emotional support and encouragement that a human trainer can offer. While an AI personal trainer may be a helpful tool for some people, it's important to consider the limitations of this technology and to seek out other sources of motivation and support if needed."

Fascinating but still early days.

Waiting for an AR interface coupled with an artificial influencer who knows you better than yourself,

 

who takes the feminine shape your like the most, the voice you like the most, the word you like the most and its marginal cost is 0.

 

That's because he/she/it analyzes your heartbeat and the micro movement of your eyes.

 

 

https://instagram.com/lilmiquela?igshid=YmMyMTA2M2Y=

Edited December 11, 2022 by Dave86ch

[UK]

Time to rewatch this movie: https://www.imdb.com/title/tt1798709/

 

[UK]

https://stratechery.com/2022/ai-homework/

[formthirteen]

ChatGPT is terrible news for accountants:

 

 

Accountants that use ChatGPT are going to be in trouble...

[LC]

This is a problem with any machine learning/ data-driven approach vs. a statistical/ mathematical approach.