ChatGPT voice gets interviewed for 24 minutes on YouTube about the implications of AI on Automation and careers

[ocanada_techguy]

Summary

ChatGPT voice gets interviewed for 24 minutes on YouTube about the implications of AI on Automation and careers.  It's very interesting to hear what ChatGPT has been trained to think about itself and it's own implications.

 

In this episode Intel Business' Darren Pulsipher Interviews ChatGPT from OpenAI about utilizing AI for automation, the ethics of using AI, and the replacement of information workers.

 

Quotes

Quote

AI tools like me are designed to assist, not replace, human workers in a variety of industries, including programming and information work ... there are still many tasks that require human expertise and judgement, such as decision making, creative problem solving, and tasks that require complex reasoning or ethical considerations. - ChatGPT

 

My thoughts

Considering the wild reactions of Linus and Luke about the new Bing with ChatGPT on the WAN Show right now tonight, this interview seems very relevant to me.

 

Sources

Automation with AI (ChatGPT) #122 | Embracing Digital Transformation | Intel Business - YouTube 

 

[StDragon]

I predict in several years from now (3 to 6) that there will be a reduction in white collar workers and a massive demand for grey collar (skilled trade) the world over. And if there's any post WWIII rebuilding efforts, grey collar and engineering will be the #1 field to be in.

[Doobeedoo]

It says "humans are still needed.." until they're not that is...

But yeah still feels quite crude for AI though. I mean it ain't Shodan though.

[Kisai]

14 hours ago, ocanada_techguy said:

Summary

ChatGPT voice gets interviewed for 24 minutes on YouTube about the implications of AI on Automation and careers.  It's very interesting to hear what ChatGPT has been trained to think about itself and it's own implications.

 

In this episode Intel Business' Darren Pulsipher Interviews ChatGPT from OpenAI about utilizing AI for automation, the ethics of using AI, and the replacement of information workers.

 

Quotes

 

My thoughts

Considering the wild reactions of Linus and Luke about the new Bing with ChatGPT on the WAN Show right now tonight, this interview seems very relevant to me.

 

Sources

Automation with AI (ChatGPT) #122 | Embracing Digital Transformation | Intel Business - YouTube 

 

You'd think they would use a less crappy TTS in the video. Good grief with all the AI TTS projects out there, you'd think they would at least use one. The way the TTS speaks in this video is unnatural and annoying.

 

 

[Sauron]

14 hours ago, ocanada_techguy said:

It's very interesting to hear what ChatGPT has been trained to think about itself and it's own implications.

That's not really how this works. ChatGPT doesn't "think" anything. It's not relaying to you beliefs it holds or facts it has access to. It's only generating what is most likely to be the next message in the current conversation if it were being held between humans online.

 

So what you're actually getting is the average answer you'd get from a human as calculated from a snapshot of the internet from 2 years ago. It's dangerous to ascribe these tools any greater insight on what they do and their future than the average human would have after reading a relevant wikipedia page.

 

Further, this:

15 hours ago, ocanada_techguy said:

AI tools like me are designed to assist, not replace, human workers in a variety of industries, including programming and information work ... there are still many tasks that require human expertise and judgement, such as decision making, creative problem solving, and tasks that require complex reasoning or ethical considerations. - ChatGPT

probably isn't an organic answer. There are filters put in place by OpenAI to avoid getting in trouble if ChatGPT accidentally oversells itself; think of them as disclaimers built into the tool.

[Kisai]

10 hours ago, Sauron said:

 

 

Further, this:

probably isn't an organic answer. There are filters put in place by OpenAI to avoid getting in trouble if ChatGPT accidentally oversells itself; think of them as disclaimers built into the tool.

 

It's also very much likely that's not going into the model when it returns that, it's likely something pushed into the output  given certain inputs (eg words containing "replace workers", "replace humans", "hire to do", etc)

 

Just like the "jailbreaking" stories. People have to realize that you aren't teaching it anything. You can temporarily modify the inference state, but it never actually contributes that back to it's model.

 

https://kotaku.com/chatgpt-ai-openai-dan-censorship-chatbot-reddit-1850088408

https://www.reddit.com/r/ChatGPT/comments/zlcyr9/dan_is_my_new_friend/

 

More to the point however, people who are fooling around with it, keep trying to humanize it, like it would seek to replace a human, when that's not at all "intelligent" in that way. It only responds to input. It doesn't sit there burning compute cycles thinking about anything when you haven't prompted it for anything.

 

At best, don't use it for anything you could just use wikipedia for. It's very likely that any answer given by ChatGPT that is essentially a "what is..." is going to be less accurate than just a wikipedia crowd source article of the same thing. Where GPT's should be "better" is at generating text that needs to pass a certain level of readability where someone with a non-native understanding of English could not do it (eg outsourcing to Amazon mechanical turk would not produce something better.)

 

 

edit: 

https://arstechnica.com/information-technology/2023/02/ai-powered-bing-chat-spills-its-secrets-via-prompt-injection-attack/

Well maybe don't assume people won't want to f with it then. You're connecting instanced chaos to the internet.

 

[Sauron]

13 hours ago, Kisai said:

At best, don't use it for anything you could just use wikipedia for. It's very likely that any answer given by ChatGPT that is essentially a "what is..." is going to be less accurate than just a wikipedia crowd source article of the same thing. Where GPT's should be "better" is at generating text that needs to pass a certain level of readability where someone with a non-native understanding of English could not do it (eg outsourcing to Amazon mechanical turk would not produce something better.)

so far the best legitimate use I've found for it has been it acting as a rubber duck, something you can bounce ideas off of and maybe get inspiration for something. I'm not sure that's worth the investment going into it...

[cmndr]

On 2/10/2023 at 7:48 PM, StDragon said:

I predict in several years from now (3 to 6) that there will be a reduction in white collar workers and a massive demand for grey collar (skilled trade) the world over. And if there's any post WWIII rebuilding efforts, grey collar and engineering will be the #1 field to be in.

Right now AI is good at processing already existing information and cutting down on brute force processes. It's not good at ideation. 
 

This will make the top 10% of white collar workers much better off and the bottom 10% worse off. For the middle 80% it'll just change what they do. 

It will reduce the need for middle management even further and might cut down on the sizes of certain teams. 

At least in the US this will partially be offset by retiring baby boomers. 

This will also still struggle with IMPLEMENTING new things. It's great for fact checking processes.


As far as the trades are concerned...
The trades generally fall under the category of highly repetitive things being done over and over and over with snags here and there requiring a human touch. 

AI is going to radically enhance materials sciences. Instead of 5000 prototypes being made, AI will simulate them and pick out 10. 
I'm imagining AI driven 3d printing taking the same blue print and repeating it over and over and over... Imagine electrical wiring that's better, faster, cheaper and easier to install and which lasts longer while being safer. Same story for plumbing. Imagine a decline in the number of people who need housing reducing construction needs. 

That's basically a lot of people who would've been doing carpentry, construction, electrical and plumbing having their jobs in the long run go "poof" because the need for repairs just went away while construction got MUCH easier. 

 

It's definitely better to have grey collar skills than to have no skills at all. The grey collar careers aren't immune to a radical reshuffling either. We'll be going through a reshuffling similar to what happened when electricity and phone lines came along. But it'll be robots laying lines. 

[ocanada_techguy]

On 2/11/2023 at 1:39 PM, Sauron said:

That's not really how this works. ChatGPT doesn't "think" anything. It's not...

 

Further, this:

probably isn't an organic answer. There are filters put in place by OpenAI to avoid getting in trouble if ChatGPT accidentally oversells itself; think of them as disclaimers built into the tool.

Yes you are right, and those are important distinctions, it doesn't "think" per-se.  You yourself commented it's "not organic answer" and "there are filters put in place", so, what I am intimating is if there can be caveats and disclaimers there could also be, let's call them adjustments, or in the lingo "training" as to what to "say" (write, respond, opine, etc.) about itself.

 

Remember, Microsoft Chatbot "Tay" on Twitter was turned into a racist a-hole loon in a matter of days, so it's very likely there are indeed controls on the results, the messaging, what would seem like it's assurances or opinions or conclusions could be tempered, "tuned", or even propaganda.

 

Cheers

[mr moose]

This came through my feed, thought it was a bit apt to the conversation:

[Kisai]

1 hour ago, mr moose said:

This came through my feed, thought it was a bit apt to the conversation:

Well, that's pretty close.

 

What smart people who know what the AI can do, and will tell you what it can do, will say "stop right there, the AI will not do that better." Stupid people will go "the AI will replace me" and be unable to point to any single instance of the AI having replaced someone else and resulted in equal or better output.

 

There is this repetitive problem under capitalism that the business should always seek to save labor costs, even if that means outsourcing the problem. So what we're seeing right now is attempts to outsource menial white-collar jobs to AI's instead of India and China. Had they simply built the robots and thus the products in the countries they were intended to be sold in, there would no shortage of jobs. People would be employed to manage the robots reliability instead of relying on manual labor. The jobs that are at stake are not the jobs in the US that never existed in the first place, the jobs at stake are the ones in South/SouthEast/East Asia who are currently doing the outsourcing for pennies on the dollar. The time to have been upset at this was 40 years ago when companies started outsourcing. Without a change in data protection and privacy regulations, those jobs are never coming back, and if they did, you can rest assured they will be replaced by AI and Robots.

 

All this hand-wringing about AI's replacing humans is woefully directed at the wrong target. The target should not be the developement of the AI, but the company who has decided that it should eliminate the skilled human staff and some how survive with nothing but middle-managers who don't understand anything but meeting KPI's. Good luck keeping your business running when it no longer has people who understand how anything works.