People are in over supply, not chips

[Arthur Hanson]

With AI/ML and advanced automation coming to everything, it is people that are in over supply. Already our environment is stressed and degrading due to over population. At the rate tasks are automating and with AI/ML getting into areas once dominated by human labor, the world has to many people and our ecosystem is becoming more distressed every day, with no signs of reversal in sight. Any thoughts or comments appreciated. Elon is right, Mars has to be our next ecosystem, maybe in a hundred years.

 

[Aaron Applebaum]

Actually, people are in under supply . The world is aging rapidly and the population bust is apparent in almost every country. More people provide new and better solutions to food supply; technology innovation; solutions for world pollution; and everything else needed to improve the standard of living. The theory that the planet cannot sustain more people has been debunked many times over.

 

[Daniel Nenni]

Hopefully AI will ease the people supply. Wife bots, husband bots, child bots, pet bots, virtual reality, all good for the environment and will trim population. Flesh and blood is over rated.

 

[Xebec]

I think it's a little more complex.

People are in undersupply for the advanced economies of the world, and also for handling the 'borrow from the future' financial situations most (all?) governments/voters/societies have put us in. The downward long term trend on population could cause serious economic collapse if not tended carefully.

OTOH there is not an infinite amount of space, resources, water, air, etc. on this planet. Ultimately quality of life is capped based on how many people are on the planet -- take all available resources and divide by the number of people, and that calculation decreases as population increases.

I think we need more people right now but long term we really need more "Earth" focus on sustainability across the board. We REALLY need something like Aliens to appear, scare us, and unify us vs. this stupid (potentially world ending) bickering we do as a species today.

 

[TBiggs]

Advancement in LLM driven AI is all currently FOMO. I have not seen any real commercial money or profit yet to end application. Till this materializes it could be a mirage , time will tell.

For companies to grow, think Apple, TSMC they need expanding markets to drive growth and enable the huge investments. Nvidia is a golden place, but who is going to pay for all this? Robots won’t it take people clawing from the bottom working hard to get a bette life. Automation, robots nor AI drive this. Will be interesting when this resets, interesting times. What will we do with a trillion comput on a chip?

 

[blueone]

Advancement in LLM driven AI is all currently FOMO.

While I can understand why you'd have that reaction, I can see targeted applications where it looks like there are revenue opportunities. I think right now they are rare, because no-cost or low-cost LLM access IMO is being used for marketing research, and industry-specific applications are nascent at best. Three very big revenue opportunities I'm familiar with are:

1. Software development, both as a coding assist and a debugging assist. I'm also seeing a lot of value at taking some poorly-written code and rewriting it to follow a style guide, which is in my experience is something a significant percentage of programmers are too lazy or unprofessional to do themselves. I've tried ChatGPT for this, and all I can say is I think there's a very significant revenue opportunity for products which are brilliantly targeted at this problem. You currently can't trust that ChatGPT code is correct.

2. The legal profession. This might be a bigger opportunity revenue-wise than software development, because programming languages and run-time environments are pretty much universal, but, just to use the US as an example, there are different document requirements for courts at the local, state, and federal levels. Law firms are very labor-intensive, they use a hierarchy of legal assistants, paralegals, intern attorneys, and attorneys at various seniority levels to deal with all of the court filing, contractual, and legal documentation requirements. If we assume the first phase is only human-directed research and providing examples, that would be a huge win for many law firms. And ChatGPT cannot be trusted in its current form, so there is a lot of value to be added. (I'm been told first-hand of an example where a hapless attorney filed a motion with a state court which quoted precedents to support the motion. The opposing attorney reviewed the proposed motion and looked up the precedents on Westlaw, and found the LLM was hallucinating about some of them. The opposing attorney reported this to the judge, and the filing attorney was forced to admit he used "software" to create the motion, was professionally embarrassed (though not sanctioned like he should have been), and had to withdraw the motion.)

3. Automated customer service bots that are effective. I'm sure we've all had the experience of these "automated assistants" when trying to get answers to customer service problems, and since I've never had a productive encounter, I assume many people have been dissatisfied too. The savings in human call center labor would be huge. Recently I used their website to open a chat session to contact Xfinity to report an equipment issue outside my home. The chat bot was profoundly ineffective, and ignored me when I asked to be connected to a representative. I finally got frustrated with the software and ended the chat session. The bot then followed me by texting me, asking if my problem was resolved and how it could help. After several useless responses I got frustrated with its ineffectiveness and its repeated texts, so I responded "Shut up". Apparently some programmer thought he or she was being funny, and the software responded to my "Shut up" with this text:

I can help you with your cancellation request, but first I'll need a bit more info. What would you like to do?
1 - Move services to new address
2 - Cancel my Xfinity services
3 - I need help with something else

 

[Paul2]

When fabs were downsizing on personel with automation, there was a huge wave of unemployment 20 years ago, I wonder, how people of that wave fared?

 

[MKWVentures]

When fabs were downsizing on personel with automation, there was a huge wave of unemployment 20 years ago, I wonder, how people of that wave fared?

there was a huge wave of unemployment? when and in what jobs exactly? People moved but it seems like we were short on being able to hire qualified people

 

[MKWVentures]

While I can understand why you'd have that reaction, I can see targeted applications where it looks like there are revenue opportunities. I think right now they are rare, because no-cost or low-cost LLM access IMO is being used for marketing research, and industry-specific applications are nascent at best. Three very big revenue opportunities I'm familiar with are:

1. Software development, both as a coding assist and a debugging assist. I'm also seeing a lot of value at taking some poorly-written code and rewriting it to follow a style guide, which is in my experience is something a significant percentage of programmers are too lazy or unprofessional to do themselves. I've tried ChatGPT for this, and all I can say is I think there's a very significant revenue opportunity for products which are brilliantly targeted at this problem. You currently can't trust that ChatGPT code is correct.

2. The legal profession. This might be a bigger opportunity revenue-wise than software development, because programming languages and run-time environments are pretty much universal, but, just to use the US as an example, there are different document requirements for courts at the local, state, and federal levels. Law firms are very labor-intensive, they use a hierarchy of legal assistants, paralegals, intern attorneys, and attorneys at various seniority levels to deal with all of the court filing, contractual, and legal documentation requirements. If we assume the first phase is only human-directed research and providing examples, that would be a huge win for many law firms. And ChatGPT cannot be trusted in its current form, so there is a lot of value to be added. (I'm been told first-hand of an example where a hapless attorney filed a motion with a state court which quoted precedents to support the motion. The opposing attorney reviewed the proposed motion and looked up the precedents on Westlaw, and found the LLM was hallucinating about some of them. The opposing attorney reported this to the judge, and the filing attorney was forced to admit he used "software" to create the motion, was professionally embarrassed (though not sanctioned like he should have been), and had to withdraw the motion.)

3. Automated customer service bots that are effective. I'm sure we've all had the experience of these "automated assistants" when trying to get answers to customer service problems, and since I've never had a productive encounter, I assume many people have been dissatisfied too. The savings in human call center labor would be huge. Recently I used their website to open a chat session to contact Xfinity to report an equipment issue outside my home. The chat bot was profoundly ineffective, and ignored me when I asked to be connected to a representative. I finally got frustrated with the software and ended the chat session. The bot then followed me by texting me, asking if my problem was resolved and how it could help. After several useless responses I got frustrated with its ineffectiveness and its repeated texts, so I responded "Shut up". Apparently some programmer thought he or she was being funny, and the software responded to my "Shut up" with this text:

Chat CPT listed Chat bots as one of the best examples of AI use. Siri and Alexa as well. If you just added root canals, it would be a complete list of my favorite things LOL

 

[blueone]

Chat CPT listed Chat bots as one of the best examples of AI use. Siri and Alexa as well. If you just added root canals, it would a complete list of my favorite things LOL

Hmmm... another ChatGPT hallucination.

 

[Daniel Nenni]

Hmmm... another ChatGPT hallucination.

The more I use Chat-GPT the more of a bubble I think it is. Still useful but must be fact checked. Siri is getting much better, I use it everyday. Alexa is annoying. We are big Amazon users and it is ad central inside my house. If it did not have the package delivery alert so we can avoid porch pirates we would turn it off.

 

[blueone]

Alexa is annoying. We are big Amazon users and it is ad central inside my house. If it did not have the package delivery alert so we can avoid porch pirates we would turn it off.

I set the option on my Amazon account to text me and send me an email when a package is delivered. No Alexa involved. I really like how Amazon sends me a photo of the delivery location, so I can figure out which of the several exterior doors our house has the Amazon delivery driver thinks is our front door.

 

[Daniel Nenni]

I set the option on my Amazon account to text me and send me an email when a package is delivered. No Alexa involved. I really like how Amazon sends me a photo of the delivery location, so I can figure out which of the several exterior doors our house has the Amazon delivery driver thinks is our front door.

Yes, my wife gets those because texts annoy me too. We have a Ring doorbell so I see deliveries coming and open the door as they arrive. They do not necessarily like that.

Porch pirates have been hitting our neighborhood pretty hard of late. We all have videos of them doing it but the police have to catch them in the act apparently. Our town has license plate readers (great AI technology) so we give the police a picture of the car w/license and they get on the watch list. If the car does not have license plates they get pulled over anyway which is how most of them are caught.

Desperate times in California, absolutely.

 

[russell.mohn]

I'm optimistic that human beings can drive our home, planet Earth, to a better state. I believe every generation embodies, however incrementally, advancements in knowledge and consciousness. It looks dour sometimes as we go about our daily lives and read the news, but seeing Earth from space (perhaps Mars) conjures different thoughts. The asymptote for the knowledge of LLMs and AI/ML is the sum total of human knowledge. And that's only human knowledge documented in language. There are domains of knowledge beyond human language. I don't believe the LLMs will go further. I've always liked Bucky Fuller's ideas, and think his World Game was way ahead of its time. He said the problems stem not from too many people consuming too many resources, but the unfair allocation of Earth's resources. https://www.bfi.org/about-fuller/big-ideas/world-game/

 

[blueone]

I've always liked Bucky Fuller's ideas, and think his World Game was way ahead of its time. He said the problems stem not from too many people consuming too many resources, but the unfair allocation of Earth's resources. https://www.bfi.org/about-fuller/big-ideas/world-game/

Unfortunately, the evidence is that the world is quickly moving farther from Fuller's altruistic view, and there's realistically nothing that we can do about it. Personally, I don't want "fair" allocation, because the definition of fair is often perverted to use it as an instrument of acquiring power. The notion has become so tiresome to me, that every time I read the words "unfair", "fair share", "equitable", and lately even "equal", I roll my eyes and figuratively walk away.

 

[count]

For the last 30 or so years until very recently manufacturing industry has had a surplus of labor because factories were generally closing faster than people were leaving the labor force. Now we have growth in factories and all of them canot seem to find people because the manufacturing workforce in the US rapidly approaching retirement age.

I am not just talking about the unskilled/semi skilled people doing manual tasks on the assembly line, but the people who design the assembly lines, program the robots, manage the flow of materials through a factory, and so on.

 

[Paul2]

For the last 30 or so years until very recently manufacturing industry has had a surplus of labor because factories were generally closing faster than people were leaving the labor force. Now we have growth in factories and all of them canot seem to find people because the manufacturing workforce in the US rapidly approaching retirement age.

I am not just talking about the unskilled/semi skilled people doing manual tasks on the assembly line, but the people who design the assembly lines, program the robots, manage the flow of materials through a factory, and so on.

Skilled assemblers with 7-10 years experience, references, some certifications, and paid training cost $2500 per month in China. These are the kind of assembly line workers you will entrust to assemble medical devices.

 

[blueone]

Skilled assemblers with 7-10 years experience, references, some certifications, and paid training cost $2500 per month in China. These are the kind of assembly line workers you will entrust to assemble medical devices.

Does the $2500/month include medical benefits?

 

[Paul2]

Does the $2500/month include medical benefits?

No, but med insurance premiums are mandatory on the employer in most provinces today. With all extra payments, the total cost of a worker is 130%-150% of net salary depending on the province.

 

[lilo777]

Advancement in LLM driven AI is all currently FOMO. I have not seen any real commercial money or profit yet to end application. Till this materializes it could be a mirage , time will tell.

For companies to grow, think Apple, TSMC they need expanding markets to drive growth and enable the huge investments. Nvidia is a golden place, but who is going to pay for all this? Robots won’t it take people clawing from the bottom working hard to get a bette life. Automation, robots nor AI drive this. Will be interesting when this resets, interesting times. What will we do with a trillion comput on a chip?

As someone who is using LLM (ChatGPT4) every day for software development, I am pretty sure that LLMs are for real. They are useful. Monetization is a different issue. I am using ChatGPT bot in Bing browser. ChatGPT (at least the way it is offered to us) is not smart enough and limits the input capacity (4000 characters in one request) to design complex programs from a spec. The way I use it is to solve specific issues in the design of a larger application. As an example (just for developing a relatively simple script/app), I can start with a simple spec and then ask for enhancements (Bing keeps the context for up to 30 consequent requests). I noticed recently (just a few days?) that Bing started displaying ads after each request/query. So the monetization has already started. OpenAI offers $20 subscriptions or flexible model priced by query. Pricing also depends on the size of the context you want to get (8K or 32K).

In short, I do not think LLMs are ready to replace actual programmers (for any non-trivial use cases). They would need to be scaled up to be able to handle more complex issues. LLMs already require a lot of resources (even for inference) so it's not clear if or how fast they will be able to scale up.