Ignorant media is not the one who installed Copilot on my computer.
Very true, but Copilot is a small part of Ai.
True. Please read the OP as “why are they pushing very small parts of AI so hard onto people who will not be using the various other parts of AI that are useful but quite different and not being pushed at people in the same way.”
(And, okay, maybe I do need writing help when typing one-handed with a cat on my lap.)
The hype headlines don’t help. They pitch ‘Ai security’ concerns as though it’s Big Mind gaining access to nuclear weapon control, but the miniscule amount of information in the articles point to copyright and intellectual property issues.
- They want you to get hooked on it. If they’re not inserting themselves into everything, then you might not use it.
- More usage = more data for improvement, to get more people hooked on it.
- The better AI gets, the better the case to use it to replace human labor. For you, this looks like convenience. For your boss, this looks like eliminating your job.
Before I get on my soapbox, #2 above is probably the main answer to the OP question. They are forcing it into everything so they can understand how you interact with it, and the data will be used to make it more compelling to use.
Onto the soapbox of labor discipline being the ultimate goal… remember, if a company believes they can replace you with a tool that does your job at 50% quality, but at 10% cost, then lowering standards becomes incredibly profitable, and they will eliminate your job to realize that profit. This is why companies are plowing truly massive amounts of money into AI.
Notice that I said “believes they can replace you”. It doesn’t much matter if the tool is significantly better or not. We saw this with the initial offshoring booms, they were wildly wasteful, but the owner class was so enticed by the prospect of cutting headcount costs that they lit money on fire to import the more compliant workers.
Eventually the pendulum swung back somewhat into more rationalized sourcing practices, but the precedent was set, workers can be disciplined by threatening them with a cheaper labor pool. This is what the AI boom is all about. It won’t last forever, but it will do some damage. The cost of some goods and services will decline slightly, the quality will decline a lot, many people will lose jobs, a few people will make ungodly sums of money.
Not sure I follow your point. The quality of Ai call centers will depend on their design, training and maintenance. Just like human call centers. The control is centralized so more people in US cities will be employed and fewer in India and US womens’ prisons.
The IDE I described above allows an expert in a field to implement computer solutions without being a programmer. But, even that is an overstatement. Many people today are engaged in programming by another name. Dealing with geographics or demographics in Arc GIS is not C++(although it can be) but it is still programming. Future updates of Arc GIS will have the prompt Ai interface, a different skill that is application oriented. That doesn’t mean a manager can handle tasks directly. It still requires those skilled in the art to interface with the Ai.
In defining products and projects there is the difference between ‘want’ and ‘need’. Management is good at stating wants, the troops in the trenches are better at meeting needs. Ai doesn’t replace them.
Modern systems have already eliminated much of the labor costs. Labor in a car is less than 10% of the cost. Ai will not significantly change it.
I have a Nobel Laureate friend who complains that a secretary used to transcribe his books from dictation. Now he is expected to write them himself on a lap top. But soon, he may do this by dictating a book into the ‘prompt’ and letting the Ai format it. Increase in productivity where the job loss has already occurred.
Sure, there are some jobs that will go the way of vacuum tube circuit design, but there are more that will be created.
I have no idea if @Crane is right that there will be more jobs created than lost, but I am pretty sure that your concern is, while understandable, misguided.
Big trend globally, and definitely in the United States, discussed here, is a declining birth rate, coupled with those who reach retirement age living longer. This country, the world, will progressively have a smaller and smaller fraction of adults actively working and producing. We will increasingly need higher productivity per active worker to keep our economies strong. Even just to keep total productivity the same per capita.
The hope is that the various forms of AI can mature to meet that demand or even increase productivity.
The concept that jobs will just be eliminated is predicated on the false assumptions of a fixed number of workers available, with a fixed amount of retirees, and a fixed number for productivity. If those were true than productivity by AI would replace workers. But none of those assumptions are valid.
I don’t think Copilot is shoved with that in mind, but that hope for increased total productivity is much more the dream of industry than is job elimination.
AI is being pushed hard in order for a few enterprises to lock up the technology and exploit it for profit. China’s announcement of lower cost open source AI created a shockwave because of the plans already in place to have public money fund these businesses for private benefit. They are pulling out the stops to eliminate competition in the market place including the calls for concern about the dangers of AI which dwarf in comparison to the dangers that already exist from controlling worldwide communications and amassing databases of private information to be exploited by both business and the state.
This seems to contain the assumption that workers can be more productive than they are. But people can’t sprint from 18 to retirement: humans are only so productive, and already US work culture is, I think, unhealthy to the point where it’s unproductive in the long run. If people are miserable and stressed out, what’s the point of a good economy?
What we can do, and where AI can help, is what we’ve done for the last 200 years (at quite some cost to the environment, but who needs plants and animals when we have economic growth): automate certain labour-intensive processes to free human workers to do what humans do better.
I sat through a lot of AI related presentations at the last American Chemical Society national meeting and really only saw two use cases. Well, maybe three, if you consider what could be a more efficient design of experiment study to count as something different. It was all either safety (given these experimental parameters and access to the general safety literature what could go wrong?) or reaction prediction (the DOE stuff mentioned or just a juiced up version of what stuff like Reaxys was doing fifteen years ago with retro synthetic analysis.) A few hours of listening to various companies either pitch a product or share internally developed case studies and all I got was that was that at best it was a souped-up search engine that might make going through the literature faster. Or again, just a fancier version of what people have been doing for years with stuff like DoE and models of all sorts.
Yes. Not by sprinting harder or longer but by having tools that increase the productivity per hour worked.
The various forms of AI are no different than other tools in this regard, just more in line with the idea that the work we want to amplify today and in the future is less manual labor, we are not replacing scribes with the printing press even, than cognitive work.
And even that amplification of cognitive work is just an acceleration of what has been happening for decades when the human computers of the early space program were supplanted by machines that did more calculations faster.
It cannot be stressed enough that demographic changes will require this sort of higher productivity going forward.
This is a bit of a hijack, but do you mean customized processor? If you do, good luck with that. I worked on that idea 50 years ago. The group I was in in grad school was working on a network of microprogrammable minicomputers which could be customized for a particular job. That idea never went anywhere. There was a company that let you customize instruction sets for processors. It was a failure. Larry Ellison bought Sun in part because he thought Sparcs could be customized to support the Oracle database. It could, but the improvment in performance was never that great.
General purpose machines have always beaten customized machines in the long run.
Now, interactive program development I can see. The best thing the AI could do is to ask about corner cases. I found that when a random engineer wanted a new feature, they only considered the typical use and almost never thought about odd inputs. A lot of bugs come from never expecting something to happen.
Agreed, that is the point. No hijack, I was addressing the broader issue of what is being hyped - the Ai markets that will be addressed. The largest market is training software that runs on Edge computers to develop software for embedded consumer and industrial applications. Probably as a plug in to the IDE. SOCs can now handle Ai software for target apps. Chips like Snapdragon have speed and obscene amounts of memory. Nothing custom. Product differentiation is obtained through the training set and board hardware.
Students anticipating programming as a career will have to be more knowledge in prompts than C++. That’s not hype. The nonsense about stand alone Ai totally replacing employees is hype.
Must have been a very exciting project at that time. Fifty years ago, wow - National and AMD had 4 bit slice micro-programmable processors. National even had an off the shelf programming system that allowed you to create the microcode then execute it by stringing together boards containing 4 slices each. I had a customer that implemented a 100 bit wide processor. For what application I never knew. You should have called me!
Oh yeah, the connection to the OP. I believe that part of the push (in addition to lust, avarice and raw greed) is to prepare the engineering world for a major paradigm change. It’s similar to the microcontroller introduction that “moved a lot of engineers from springs, levers and gears to software and semiconductors” (Floyd Kvame).
Was there a presentation on Microsoft’s MatterGen, the novel materials exploration tool? They tried to sell us on it, but it was apparent that the very senior materials researcher they paired with it was equally important and don’t have one of those.
There might have been, I was mostly going to those presentations when there wasn’t something I was more interested in. My take on the whole thing is that I’ve lived through a number of fads that never really amounted to anything. I could see AI as a more powerful way to handle cheminformatics and maybe to help guide a faster DOE-like experiment system. It’s not like AI can actually go into the lab and see if the idea worked.
This is just the nobler social-science rationalization of what’s always been true - the owner class is paranoid about keeping wages under control, because higher wages are not only a menace to profit margin, but they risk transferring more power from the owner class to the working class.
Owners have always wanted to squeeze more productivity out of workers, because this provides more margin to to skim the fruits of their labor in the form of profits.
I don’t dispute the demographic reality that birth rates are getting tighter, not at all. Nor that this could result in higher prices. The tell here is that the owner class never has any solutions for making it easier for workers to grow family, such as generous child-leave policies, social safety nets, ever. Instead we get pressure to eliminate reproductive choice and force more births, to create wage scarcity that mandates 2-income familiies, to increase the efficiency of labor extraction in ways that provide the macroeconomic benefits of higher birthrates, while robbing every ounce of joy that’s natural to having a family. Deep down they don’t want more workers, because their ideal social relationship is a working class that’s highly productive and exploitable, but in small enough numbers that they pose no real threat to the owning class.
So that’s where all the fake anxiety is coming from about the falling birthrate. It’s just a different form of the perennial fear of workers gaining more leverage, and this is why it will not stop after the productivity gains offset any labor market tightness.
We know this because the productivity obsession predated birth rate concerns! It’s decades old, if not centuries old. The AI boom started during the Great Recession when the labor market was extraordinarily loose for long periods of time! It’s simply a matter of coincidence that it happened to come to fruition during the COVID era. But the absolute mania about it is definitely because it came to fruition in a time when capitalism was having a five-alarm fire drill about highly constrained, “undisciplined” labor markets.
AI is at root about putting workers in their place. To speak to the OP, it’s becoming ubiquitous so that it can suck up every detail about how workers do their jobs in order to replace them, and to start acclimating them to this situation becoming normal. They want you to start exploiting it so that you’ll accept it as normal when it starts exploiting you.
Let’s be very clear. Every individual is going to be interested in maximizing their own power which means wealth. Owners of companies (which includes all of us invested in the market even in our retirement funds) want our companies appreciating and/or returning dividends and making us more wealth. And those not invested want employment and high wages. Most are both workers and owners to some degree, and those who are not at all owners generally have their wages and employment best served by a growing economy.
Total productivity increasing more than total wages is the recipe for more profit. My WAG remains that there is more salivating over increasing productivity by AI than over decreasing real wages.
The makers of the AI want to sell both workers and more so owners that their tools (and not the competitors’ tools) are the way to increase productivity. Workers hoping that will mean fewer hours or more pay (it won’t) and owners hoping it means more productivity per unit wage. Pushing it on consumers is part of trying to make that sale.
No. Some of us are not interested in maximizing power or wealth. Sometimes, people stop when they have enough. Greed is not a virtue.
Specifically, all executive classes whether in business, government or elsewhere want to maintain as much of a monopoly on agency as they can. They want to be the planners and the shot-callers. They view a workforce with its own agenda with about the same favor as a general views privates who question their orders.
This was all part of my PhD research. I designed and implemented an object oriented microprogramming language. And when I graduated in 1980 I looked around and guessed that the field was dying so I had better switch, which I did. And I was right, though I wrote some papers and book chapters while working on other things.
The level of customization you mention sounds reasonable. Special purpose processors still pop up from time to time, not so much now that they are way too expensive when you can buy an ARM core and customize the surrounding logic.