Bullshit code as a rule doesn't compile, and if it does, it cannot be debugged;
To date the most advanced code generated apps are one button apps that say 'change color on box'
The latest bullshit so called 'Function Calling' is just JSON, where the human has to write the JSON so that the GPT knows what to generate and the idea is the query will fill in the fields, and then a new JSON is generated that is used as an argument call to a function.
Like all bullshit, what they're doing is making up new bullshit to force the bullshit that doesn't work, to work on special cases of bullshit;
The economic viability of a 4-hour work week is a contentious issue. As per a lecture from MIT, historical trends indicate that improvements in efficiency and productivity have not resulted in a decrease in the workweek. Instead, they have led to an increase in wages and per capita consumption. This implies that a substantial transition to a 4-hour work week may not be economically viable within the existing capitalist framework, where the pursuit of profit frequently takes precedence over the desire for reduced work hours. Nonetheless, there have been experiments with part-time roles and shorter workdays, suggesting that such a model could potentially be adopted under specific circumstances.
The introduction of AI, particularly Generative AI, is often discussed as a potential game-changer in this context. However, whether it can serve as a panacea for the challenges associated with reduced work hours remains a topic of debate. The concept of Universal Basic Income (UBI) often emerges in these discussions as a potential part of the solution. Critics of UBI argue that it could be prohibitively expensive and could potentially exacerbate poverty. On the other hand, proponents contend that a basic income could help lift millions from poverty and provide a crucial financial safety net.
Thanks for your insights! I agree it's not an easy problem at all. However, what Graeber argues is that the existence of bullshit jobs is an "intentional" attempt to stop the possibility of working less.
This is a great title that chatGPT could never write haha.
GPT is challenging our entire society. I find it amusing how much uproar, attention, and discomfort it is making us feel as a whole. It definitely isn't Einstein in a genie bottle, but it's the first interface to the cutting edge of AI that the general public has seen. This specifically was made to start the conversation of AGI in public discourse for whatever reason you want to believe. I am not here to debate that issue. I remember when people use to laugh about AGI and now it's being taken seriously. This is a fundamental shift in our society whether it will come sooner or later.
In terms of coding, GPT-4 is definitely not replacing any jobs. But it has brought access of a skill to more people. Yea, it will write bad code, but all of us have written bad code at some point. The point is that people can start and fail and learn quicker. This is powerful. The point of technological advancement is to empower people and bring advanced tools to more and more people. For example, programming industrial robot arms use to require a Masters degree but now a technician with some YouTube videos can program an arm to do a task.
GPT-4 isn't a tool that rids us of all the work, but it is also not a complete bullshit generator. It's a stepping stone in the technological arc. The Transformer architecture has been a convergence moment to approach data ingestion and regurgitation for computers in a semi-elegant way. And now, the investment dollar is here and every top level AI Engineer has a bag full of cash to tackle AGI.
You can doubt GPT-4's intelligence, but let's not doubt the intelligence of our most brillant minds all focused on bringing AI to the next frontier.
I dislike the idea of “bullshit jobs”, and maybe as someone who spent a decade in Big4/large corporation, I’m biased.
Large organizations are all about information flows and decision making. That’s what many “bullshit” jobs are - you get several pieces of information, combine them, make certain decisions based on your expertise, and pass them along.
The fact that many people don’t like their jobs doesn’t automatically mean that they are not necessary.
Now, current tools like ChatGPT can greatly help such people, making information processing and decision-making much quicker and easier. However, the modern organization still expects the person, not the machine, to do it.
I think of it as if every “bullshit worker” suddenly got an employee that can do 80% of what the worker used to do. The worker becomes a manager of bullshit employees, and being a manager of someone who can work 24/7, doesn’t want a raise, and never takes sick leave is a much more fun job than being a real manager or being a “bullshit worker.”
In the past two weeks, I had around 20 conversations with early ChatGPT adopters in advisory and corporations. One story was particularly illustrative - the guy was overloaded with work and asked his boss to hire a junior to help. It took a few months for the boss to approve, but in the meantime, ChatGPT was released, and the guy decided he didn’t need human help anymore.
Agreed. "Bullshit job" doesn't include all jobs that *look like* bullshit jobs. There are many of those that happen to be quite critical.
Another question would be if the world could have developed during the last 100 years to do those jobs differently and allow people to have more free time (which is pretty much what moved Graeber to think of this).
So, it's not that bullshit jobs are unnecessary, it's that they have little social value—that they exist more to fill up time than anything else, even if now you can't simply remove those and expect the world to still work.
It smells like a conspiracy, to be honest. I don't think there is an evil ruling class that intentionally created these jobs to keep people busy and not revolt.
Businesses try to be as effective as possible, reduce costs and maximize profits in the long term. If they can automate/eliminate certain tasks and jobs, they will, and I've seen this happening many times. But there are certain limitations, like inertia, the complexity of large organizations, employee turnover, poor processes and documentation, and questionable workforce quality.
Additionally, governments continuously produce new compliance requirements. There are so many rules that a significant portion of corporations is working just to ensure their organization is compliant. This applies to a big portion of consulting business and most part of government bureaucracy, too.
So maybe the real "ruling class" that generates bullshit jobs is government, but I also doubt it. It's just that the compliance burden historically was not a KPI for them.
Well, we disagree on this. Not everything that smells like a conspiracy is false by default. What we agree on is that the ruling class, whatever that is, isn't necessarily evil.
4chan has proven that conspiracy turns out to be true 90% of the time
While posts on twitter&facebook are false 90% of the time;
To date the most reliable truth AI is that trained on 4chan CORPUS
chatGPT is trained on facefuck&twitter which is why its 100% bullshit generating; Bullshit in, bullshit out; If you want a real AI, run an AI using 4chan weights at home and get some truth;
Agreed if your a bull-shit generator, like the gov guy above who has to gen N pages of bullshit text a day that nobody ever reads, then chat-GPT is a gift from heaven;
The problem of course is, that as this USA economy & USD collapse, these non-productive jobs of generating text that nobody reads will go away and quick;
Like Jim Rogers says want to survive the future?? Become a farmer :)
Does such BS work have a role in the modern world or could we dispense with it to free up resources and productivity? It's reminiscent of the red tape and "fake" work in old Soviet Russia, but one would imagine this wouldn't fly in a competitive capitalistic economy.
Graeber says exactly this in his essay. And I agree. I don't have the answer to why it happens; he says this:
"The answer clearly isn't economic: it's moral and political. The ruling class has figured out that a happy and productive population with free time on their hands is a mortal danger (think of what started to happen when this even began to be approximated in the '60s). And, on the other hand, the feeling that work is a moral value in itself, and that anyone not willing to submit themselves to some kind of intense work discipline for most of their waking hours deserves nothing, is extraordinarily convenient for them."
The problem is that your problem is a western problem, where most of you are already laying on the sidewalks in your own feces drugged out, very few of you are computational scientists;
...
In my eyes say ASIA, or most of the real world, where people grow all their own food, most are quite busy sun up to sun down farming and fixing stuff and taking care of the kids, stuff that humans do; In my life of travel I would say 90% live this way;
But in say the USA they have created this world where people moved to the citys, and then they off shored all the jobs to china long ago, and now you have people bored out of their minds and doing drugs and dying on sidewalks in their own shit;
...
At least in CHINA everybody is still only one generation away from the farm, all could go back to the farm tomorrow and it would be like nothing ever changed, but in USA the GOV during the great depression destroyed the family farm, so now the city zombies have NOTHING to return to .
I think in five generations the China-ASIA family farm will still be there, because the city elite like to go home on holiday, and lots of people prefer the slow simple life, so the family farm & house never goes away,
I’ve always found the term ‘bullshit jobs’ a little disrespectful, so I wonder if you could reflect on that?
Also I feel the analogy of ‘bullshit generator’ is missing the point with regards to what makes tools like ChatGPT and Github CoPilot useful to people.
Github Copilot is incredibly effective because it suggests relevant autocompletes often enough to be generally useful in supporting during daily coding tasks.
The word relevant is key here. In the same way ChatGPT can be useful not as a database of facts, but as a tool to speed up writing tasks and present relevant suggestions often enough to be useful to people.
While organizational bullshit indeed lends itself very well for that, I find it slightly reductive of you to say that’s ChatGPT’s single purpose.
Of course. I think the problem with the word "bullshit" is that its common meaning is slightly different (and more derogatory/disrespectful) than its philosophical one, which I'm using here (https://en.m.wikipedia.org/wiki/On_Bullshit). The terms "bullshit jobs" and "bullshit generators" both take that second meaning.
I’m familiar with the work On Bullshitt, but even going by the philosophical meaning it is not free from sentiment, as he conclude that bullshitters are more insidious and “more of a threat against the truth than are liars.”
ChatGPT is more of a threat to truth than if it were a liar. I do believe that. There's no variable you can measure that will reveal whether a given statement is true or not. That's why confabulation is so problematic.
I think Graeber's definition of bullshit jobs is slightly different than Frankfurt's (see James' comment) and much softer. He's simply pointing out the pointlessness of the jobs.
Right! But what if it lied according to a pattern that you could observe and measure? Confabulation is arbitrary.
However, I do agree if you mean a tool that could *intentionally* decide when to lie and when to not lie. That's definitely more problematic than confabulating. Although I don't think we're there yet, thankfully!
Yes! But one thing is a machine that happens to lie sometimes and not some other times according to an external variable (something out of its control) and another is an entity that can decide by whatever inscrutable mechanism when to lie and when not to—without us being able to tell.
All bullshit is like all feces, you know it when you see it, you can smell it, and nobody can say otherwise, you know shit, we all know shit
In the computer biz the ECS - Emperor has no Clothes is a constant thing, but don't worry since 1940's AI has come&gone a dozen times and died of hype and this homo-llm variant being sold today by open-ai is the worst incarnation to date;
He’s referring to ‘bullshit jobs’ in the way anthropologist David Graeber coined it, and who defined it in terms of workers who classify their own work in this way. That’s hardly disrespectful.
There are dozens of examples that most people are incompetent fools at their job, that they rise to their level of incompetence and generate bullshit;
Most people are fools at work, and they are led by fools;
...
This is why anybody with a brain, either drops out of college ASAP, or in his first job at Google quits in a year and starts his own company, and one person can do incredible shit and bring a real product to market, but then as soon as you start bring in employees then you screw the pooch; The only scaling success is like early Intel where they could make employees 'millionaires' and incentive the morons to retire & quit
I think we all know in AI that the SV, or VC drive the bullshit, the dumb money that floods the startups, so called 'unicorns' which of course is just a nice word for homo, those who spend VC money on sushi&sodomy, and write apps like "Linkd'd" say sam-altman fame; Then atman was lucky enough that peter-thiel flaming CIA homo created open-ai and put ycombinator(sv homo grooming op) in power over open-ai, now sold to microsoft;
Then somebody call's bullshit, and they say you can't call open-ai 'bullshit' that's homo-phoboic, ... never ending bullshit
> I find it slightly reductive of you to say that’s ChatGPT’s single purpose.
I knew instantly upon reading the title that the author was making no single-purpose indictment of ChatGPT. In the body, there are many passages that make it clear, in the context of pointless corporate sausage-making, this tool has been ideal.
> Github Copilot is incredibly effective ...
Yes - because it's a co-pilot. ChatGPT is at best, a demo. It has no fitness-of-purpose like a co-pilot. Untouched, untamed, and in the wild, ChatGPT is perfect AS IS for some work.
co-pilot send every line of code that you will ever type home to GOOGLE for code-training, ergo they will own your algo
anybody that is stupid enough to code and give their code away for free to google, is well a fool
...
chat-GPT is a LIE, that's why its BULLSHIT
open-AI is not open, and the LLM tech is not AI, its 20 year old look-ahead mapping using recursion algos to minimize a matrix, at best you could argue that the novel 30B, 70B & 300B is only possible with cheap A100 class Tesla NVIDIA machines allow TB matrixes, otherwise we were all doing this stuff +15 years ago, but we called it LSTM, sure the 'prompt' is a new paradigm, but IMHO things have gotten worse the stuff we were running +10 years ago could do math, the new chatGPT can't even do binary arithmetic, let alone multiplication, ... Bullshit
LLM will never be self-aware or AGI, impossible, its just a random number generator that has an optimized solution to generate human preferable word orders
The training for chatGPT being facebook&twitter means you have a chat-bot with sub 80 IQ, and that all you will ever have; is a chat bot;
Lots of people on hugginface are figuring out that you can train with real books say feynman lecture on physics and actually have something a real expert, using small 7B models that you can train on a home GPU
>... co-pilot send every line of code that you will ever type home to GOOGLE for code-training
To be clear, I wasn't defending co-pilots; simply pointing out that stark difference between things that are designed to serve specific purposes and the giant ball of BS that ChatGPT uses to make everyone feel empowered.
>... its just a random number generator that has an optimized solution to generate human preferable word orders
And they wonder why all the hallucinations. There are no hallucinations. There are only very fleeting moments of clarity that may as well be random.
Nice article. Thanks. Also, are we therefore ushering in the most boring of dystopias, wherein our workplace information systems become maximally saturated with content that no person wanted to be asked to produce, that no person actually produced, and that, because it wasn’t really necessary in the first place, that no person actually reads.
yep, its near just today Zerohedge said that retail is dumping their AI stocks and that hedge-funds will follow next week, the end is near
The deal is open-ai spent 10's of millons of USD on influencers that past year on youtube to hype this bitch, and the problem is 'wheres the beef', as in there is no beef
I wonder how this bullshit-elimination effect will play out when LLMs truly become mainstream? And ironically, it's the push by Google, Microsoft, and other giants working to embed AI into every business product for the sake of productivity gains that is the most likely to get us there.
Agreed! I like this quote by scientist Roy Amara: "we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run." That's what I think will happen with these tools, too.
It's not much better than the LSTM we had in 2017, the only thing we have now is 64GB mobile-device, so in theory I can run a LLM on my phone now, but the text-gen is really not much better, I can compare I still have my live code side-by-side on my system running the same gpus the guy who did it all Karpathy is doing videos now explaining how the shit works, but the thing is he just has a bigger audience now but the tech is same-same
"Attention is the only thing you need" is six years old, there is nothing new under the sun.
Also, can't remember if I mentioned this to you, but James Suzman's book "Work" is very good, and in it he talks about the term "bullshit jobs" explicitly. I really enjoyed it.
Haven't thought of this aspect, very interesting. I suppose, as the ruling class sees free time as a danger, the use of ChatGPT to do pointless work will see as a danger too because it frees time for the employees. ChatGPT bans will ensue. Or the creation of additional tasks to compensate for the freed time resulting from the use of it.
Great article! It has many multifaceted dimensions. I think one more facet may be that so very many humans find AI threatening precisely because it can clearly show that what they do and what they’ve been up to for years is bullshit, and perhaps it always likely was. Yet even more damaging is the hugely decisive reveal linking their personal identities to bullshit, meaning precisely that the foundation of their identities are also bullish. That is a hard pill for a lot of people to swallow. It’s a real threat for people insecure in their core identities. Yet bullshit does pervade so much of human life.
One of the biggest bullshit jobs I ever saw was when I was an airline pilot flying in and out of China. As we would come through the entry doors of a major metropolitan airport there was one guy directing us to place our luggage in the metal detector while he scanned us with a wand. But that wasn’t the big bullshit job. After we passed through the wand guy and collected our luggage we had to stand in a line behind a tape. There was another guy to stop us at the tape and let the line build up. Then he would let us go. So, the essence of the job was to stop us. Then let us go. Stop us. Then let us go. Rinse and repeat. All -- day -- long. (I occasionally pondered how he would reply to his wife asking him if he had a good day at work that day - how would you reply? “Another great day honey. You should’ve seen me operate. Would you do me a favor? Just, kill me now”). In China these are called “rice bowl jobs”. They give people a sense of dignity where they get to wear uniform and do something invented which is just short of doing nothing at all. They get just enough of a wage to afford “rice bowl” of food. Such jobs are the epicenter of bullshit, but they give people a sense of dignity (supposedly) and keeps people from becoming disgruntled in an authoritarian regime building into an unmanageable mob. A lot of humans seem to crave a dignified ….. I guess you’d call it a ... Job. But I digress.
Another serious purpose I see for GPT-4, or ChatGPT, is to elevate true mediocrity to a more refined level. My 88 year old father recently delivered a reasonably good eulogy for his 94-year-old friend. It was above average by most standards, as he had been preparing it in his mind for years. When it came time to write his tribute for the local paper, I have to say it was pretty mediocre work. He’s an educated man with a Master’s degree in education, and I’m sure he thinks he’s an above average writer. But truth be told he’s a below average writer. This tribute was okay in that the sentiment he delivered was clearly sincere and genuine. But he just lacked certain skills to do a very great job. So, I volunteered to take his tribute and amp it up with GPT-4. I took my father’s contribution and applied the following prompt:
Please improve the writing of the following tribute: REMEMBERING DURAND BLANDING by Norm Toensing.
In with the marginal - out with the pretty damn impressive. The difference in quality was night and day. My father thought it was now a superb piece of writing and was incredibly appreciative. After proofreading it he only altered one word. My contribution took about five minutes. Yet in our small town I suspect my father’s reputation as an eloquent writer was secured. So, this tool also does have the ability to elevate mediocrity.
What % of ChatGPT users do you think are using it for “bullshitting” purposes?
Do you think there are other common use cases?
That's a very good question. I wouldn't dare take a guess, but probably more than I thought before writing this article. I liked this reflection by F. Chollet: https://twitter.com/fchollet/status/1672655081060319232
I'd say coding is another typical use case. I hope searching for information/facts isn't one!
Bullshit code as a rule doesn't compile, and if it does, it cannot be debugged;
To date the most advanced code generated apps are one button apps that say 'change color on box'
The latest bullshit so called 'Function Calling' is just JSON, where the human has to write the JSON so that the GPT knows what to generate and the idea is the query will fill in the fields, and then a new JSON is generated that is used as an argument call to a function.
Like all bullshit, what they're doing is making up new bullshit to force the bullshit that doesn't work, to work on special cases of bullshit;
The economic viability of a 4-hour work week is a contentious issue. As per a lecture from MIT, historical trends indicate that improvements in efficiency and productivity have not resulted in a decrease in the workweek. Instead, they have led to an increase in wages and per capita consumption. This implies that a substantial transition to a 4-hour work week may not be economically viable within the existing capitalist framework, where the pursuit of profit frequently takes precedence over the desire for reduced work hours. Nonetheless, there have been experiments with part-time roles and shorter workdays, suggesting that such a model could potentially be adopted under specific circumstances.
The introduction of AI, particularly Generative AI, is often discussed as a potential game-changer in this context. However, whether it can serve as a panacea for the challenges associated with reduced work hours remains a topic of debate. The concept of Universal Basic Income (UBI) often emerges in these discussions as a potential part of the solution. Critics of UBI argue that it could be prohibitively expensive and could potentially exacerbate poverty. On the other hand, proponents contend that a basic income could help lift millions from poverty and provide a crucial financial safety net.
Thanks for your insights! I agree it's not an easy problem at all. However, what Graeber argues is that the existence of bullshit jobs is an "intentional" attempt to stop the possibility of working less.
This is a great title that chatGPT could never write haha.
GPT is challenging our entire society. I find it amusing how much uproar, attention, and discomfort it is making us feel as a whole. It definitely isn't Einstein in a genie bottle, but it's the first interface to the cutting edge of AI that the general public has seen. This specifically was made to start the conversation of AGI in public discourse for whatever reason you want to believe. I am not here to debate that issue. I remember when people use to laugh about AGI and now it's being taken seriously. This is a fundamental shift in our society whether it will come sooner or later.
In terms of coding, GPT-4 is definitely not replacing any jobs. But it has brought access of a skill to more people. Yea, it will write bad code, but all of us have written bad code at some point. The point is that people can start and fail and learn quicker. This is powerful. The point of technological advancement is to empower people and bring advanced tools to more and more people. For example, programming industrial robot arms use to require a Masters degree but now a technician with some YouTube videos can program an arm to do a task.
GPT-4 isn't a tool that rids us of all the work, but it is also not a complete bullshit generator. It's a stepping stone in the technological arc. The Transformer architecture has been a convergence moment to approach data ingestion and regurgitation for computers in a semi-elegant way. And now, the investment dollar is here and every top level AI Engineer has a bag full of cash to tackle AGI.
You can doubt GPT-4's intelligence, but let's not doubt the intelligence of our most brillant minds all focused on bringing AI to the next frontier.
I dislike the idea of “bullshit jobs”, and maybe as someone who spent a decade in Big4/large corporation, I’m biased.
Large organizations are all about information flows and decision making. That’s what many “bullshit” jobs are - you get several pieces of information, combine them, make certain decisions based on your expertise, and pass them along.
The fact that many people don’t like their jobs doesn’t automatically mean that they are not necessary.
Now, current tools like ChatGPT can greatly help such people, making information processing and decision-making much quicker and easier. However, the modern organization still expects the person, not the machine, to do it.
I think of it as if every “bullshit worker” suddenly got an employee that can do 80% of what the worker used to do. The worker becomes a manager of bullshit employees, and being a manager of someone who can work 24/7, doesn’t want a raise, and never takes sick leave is a much more fun job than being a real manager or being a “bullshit worker.”
In the past two weeks, I had around 20 conversations with early ChatGPT adopters in advisory and corporations. One story was particularly illustrative - the guy was overloaded with work and asked his boss to hire a junior to help. It took a few months for the boss to approve, but in the meantime, ChatGPT was released, and the guy decided he didn’t need human help anymore.
Agreed. "Bullshit job" doesn't include all jobs that *look like* bullshit jobs. There are many of those that happen to be quite critical.
Another question would be if the world could have developed during the last 100 years to do those jobs differently and allow people to have more free time (which is pretty much what moved Graeber to think of this).
So, it's not that bullshit jobs are unnecessary, it's that they have little social value—that they exist more to fill up time than anything else, even if now you can't simply remove those and expect the world to still work.
It smells like a conspiracy, to be honest. I don't think there is an evil ruling class that intentionally created these jobs to keep people busy and not revolt.
Businesses try to be as effective as possible, reduce costs and maximize profits in the long term. If they can automate/eliminate certain tasks and jobs, they will, and I've seen this happening many times. But there are certain limitations, like inertia, the complexity of large organizations, employee turnover, poor processes and documentation, and questionable workforce quality.
Additionally, governments continuously produce new compliance requirements. There are so many rules that a significant portion of corporations is working just to ensure their organization is compliant. This applies to a big portion of consulting business and most part of government bureaucracy, too.
So maybe the real "ruling class" that generates bullshit jobs is government, but I also doubt it. It's just that the compliance burden historically was not a KPI for them.
Well, we disagree on this. Not everything that smells like a conspiracy is false by default. What we agree on is that the ruling class, whatever that is, isn't necessarily evil.
4chan has proven that conspiracy turns out to be true 90% of the time
While posts on twitter&facebook are false 90% of the time;
To date the most reliable truth AI is that trained on 4chan CORPUS
chatGPT is trained on facefuck&twitter which is why its 100% bullshit generating; Bullshit in, bullshit out; If you want a real AI, run an AI using 4chan weights at home and get some truth;
Agreed if your a bull-shit generator, like the gov guy above who has to gen N pages of bullshit text a day that nobody ever reads, then chat-GPT is a gift from heaven;
The problem of course is, that as this USA economy & USD collapse, these non-productive jobs of generating text that nobody reads will go away and quick;
Like Jim Rogers says want to survive the future?? Become a farmer :)
Does such BS work have a role in the modern world or could we dispense with it to free up resources and productivity? It's reminiscent of the red tape and "fake" work in old Soviet Russia, but one would imagine this wouldn't fly in a competitive capitalistic economy.
Graeber says exactly this in his essay. And I agree. I don't have the answer to why it happens; he says this:
"The answer clearly isn't economic: it's moral and political. The ruling class has figured out that a happy and productive population with free time on their hands is a mortal danger (think of what started to happen when this even began to be approximated in the '60s). And, on the other hand, the feeling that work is a moral value in itself, and that anyone not willing to submit themselves to some kind of intense work discipline for most of their waking hours deserves nothing, is extraordinarily convenient for them."
Interesting.
The problem is that your problem is a western problem, where most of you are already laying on the sidewalks in your own feces drugged out, very few of you are computational scientists;
...
In my eyes say ASIA, or most of the real world, where people grow all their own food, most are quite busy sun up to sun down farming and fixing stuff and taking care of the kids, stuff that humans do; In my life of travel I would say 90% live this way;
But in say the USA they have created this world where people moved to the citys, and then they off shored all the jobs to china long ago, and now you have people bored out of their minds and doing drugs and dying on sidewalks in their own shit;
...
At least in CHINA everybody is still only one generation away from the farm, all could go back to the farm tomorrow and it would be like nothing ever changed, but in USA the GOV during the great depression destroyed the family farm, so now the city zombies have NOTHING to return to .
I think in five generations the China-ASIA family farm will still be there, because the city elite like to go home on holiday, and lots of people prefer the slow simple life, so the family farm & house never goes away,
I’ve always found the term ‘bullshit jobs’ a little disrespectful, so I wonder if you could reflect on that?
Also I feel the analogy of ‘bullshit generator’ is missing the point with regards to what makes tools like ChatGPT and Github CoPilot useful to people.
Github Copilot is incredibly effective because it suggests relevant autocompletes often enough to be generally useful in supporting during daily coding tasks.
The word relevant is key here. In the same way ChatGPT can be useful not as a database of facts, but as a tool to speed up writing tasks and present relevant suggestions often enough to be useful to people.
While organizational bullshit indeed lends itself very well for that, I find it slightly reductive of you to say that’s ChatGPT’s single purpose.
I’m happy to be convinced otherwise, of course.
Of course. I think the problem with the word "bullshit" is that its common meaning is slightly different (and more derogatory/disrespectful) than its philosophical one, which I'm using here (https://en.m.wikipedia.org/wiki/On_Bullshit). The terms "bullshit jobs" and "bullshit generators" both take that second meaning.
I’m familiar with the work On Bullshitt, but even going by the philosophical meaning it is not free from sentiment, as he conclude that bullshitters are more insidious and “more of a threat against the truth than are liars.”
ChatGPT is more of a threat to truth than if it were a liar. I do believe that. There's no variable you can measure that will reveal whether a given statement is true or not. That's why confabulation is so problematic.
I think Graeber's definition of bullshit jobs is slightly different than Frankfurt's (see James' comment) and much softer. He's simply pointing out the pointlessness of the jobs.
Oh, I would be ten times more concerned with a tool that could lie than a tool that’s inaccurate haha. But that’s a disccusion for another time.
Right! But what if it lied according to a pattern that you could observe and measure? Confabulation is arbitrary.
However, I do agree if you mean a tool that could *intentionally* decide when to lie and when to not lie. That's definitely more problematic than confabulating. Although I don't think we're there yet, thankfully!
A lie is always intentional, isn’t it? Otherwise it wouldn’t be lying ;)
Yes! But one thing is a machine that happens to lie sometimes and not some other times according to an external variable (something out of its control) and another is an entity that can decide by whatever inscrutable mechanism when to lie and when not to—without us being able to tell.
It's not only Bullshit, its EVIL
All bullshit is like all feces, you know it when you see it, you can smell it, and nobody can say otherwise, you know shit, we all know shit
In the computer biz the ECS - Emperor has no Clothes is a constant thing, but don't worry since 1940's AI has come&gone a dozen times and died of hype and this homo-llm variant being sold today by open-ai is the worst incarnation to date;
EVIL
https://bilbobitch.substack.com/p/generative-ai-the-new-hollywood-images
Operation-Paperclip Legacy of chat-GPT
https://bilbobitch.substack.com/p/chaoschatbot-how-did-global-homo
He’s referring to ‘bullshit jobs’ in the way anthropologist David Graeber coined it, and who defined it in terms of workers who classify their own work in this way. That’s hardly disrespectful.
Exactly, thanks James. It's an adjective that describes the pointlessness of the tasks.
Peter Principal
Druckers Law
Parkinsons law
There are dozens of examples that most people are incompetent fools at their job, that they rise to their level of incompetence and generate bullshit;
Most people are fools at work, and they are led by fools;
...
This is why anybody with a brain, either drops out of college ASAP, or in his first job at Google quits in a year and starts his own company, and one person can do incredible shit and bring a real product to market, but then as soon as you start bring in employees then you screw the pooch; The only scaling success is like early Intel where they could make employees 'millionaires' and incentive the morons to retire & quit
I think we all know in AI that the SV, or VC drive the bullshit, the dumb money that floods the startups, so called 'unicorns' which of course is just a nice word for homo, those who spend VC money on sushi&sodomy, and write apps like "Linkd'd" say sam-altman fame; Then atman was lucky enough that peter-thiel flaming CIA homo created open-ai and put ycombinator(sv homo grooming op) in power over open-ai, now sold to microsoft;
Then somebody call's bullshit, and they say you can't call open-ai 'bullshit' that's homo-phoboic, ... never ending bullshit
> I find it slightly reductive of you to say that’s ChatGPT’s single purpose.
I knew instantly upon reading the title that the author was making no single-purpose indictment of ChatGPT. In the body, there are many passages that make it clear, in the context of pointless corporate sausage-making, this tool has been ideal.
> Github Copilot is incredibly effective ...
Yes - because it's a co-pilot. ChatGPT is at best, a demo. It has no fitness-of-purpose like a co-pilot. Untouched, untamed, and in the wild, ChatGPT is perfect AS IS for some work.
co-pilot send every line of code that you will ever type home to GOOGLE for code-training, ergo they will own your algo
anybody that is stupid enough to code and give their code away for free to google, is well a fool
...
chat-GPT is a LIE, that's why its BULLSHIT
open-AI is not open, and the LLM tech is not AI, its 20 year old look-ahead mapping using recursion algos to minimize a matrix, at best you could argue that the novel 30B, 70B & 300B is only possible with cheap A100 class Tesla NVIDIA machines allow TB matrixes, otherwise we were all doing this stuff +15 years ago, but we called it LSTM, sure the 'prompt' is a new paradigm, but IMHO things have gotten worse the stuff we were running +10 years ago could do math, the new chatGPT can't even do binary arithmetic, let alone multiplication, ... Bullshit
LLM will never be self-aware or AGI, impossible, its just a random number generator that has an optimized solution to generate human preferable word orders
The training for chatGPT being facebook&twitter means you have a chat-bot with sub 80 IQ, and that all you will ever have; is a chat bot;
Lots of people on hugginface are figuring out that you can train with real books say feynman lecture on physics and actually have something a real expert, using small 7B models that you can train on a home GPU
>... co-pilot send every line of code that you will ever type home to GOOGLE for code-training
To be clear, I wasn't defending co-pilots; simply pointing out that stark difference between things that are designed to serve specific purposes and the giant ball of BS that ChatGPT uses to make everyone feel empowered.
>... its just a random number generator that has an optimized solution to generate human preferable word orders
And they wonder why all the hallucinations. There are no hallucinations. There are only very fleeting moments of clarity that may as well be random.
Nice article. Thanks. Also, are we therefore ushering in the most boring of dystopias, wherein our workplace information systems become maximally saturated with content that no person wanted to be asked to produce, that no person actually produced, and that, because it wasn’t really necessary in the first place, that no person actually reads.
I think it's a likely—and depressing—possibility... The boring AI apocalypse, right: https://clivethompson.medium.com/the-boring-apocalypse-of-todays-ai-6365345444a8
yep, its near just today Zerohedge said that retail is dumping their AI stocks and that hedge-funds will follow next week, the end is near
The deal is open-ai spent 10's of millons of USD on influencers that past year on youtube to hype this bitch, and the problem is 'wheres the beef', as in there is no beef
Ooh thanks for the link.
One gets the feeling from reading Hofstader's article that he lives in a bubble, as intellectuals are wont to do.
Yes, he does. And I think I did, too. It's hard to do perspective taking sometimes!
Interesting take!
At the same time, according to recent Pew Research study (https://www.pewresearch.org/short-reads/2023/05/24/a-majority-of-americans-have-heard-of-chatgpt-but-few-have-tried-it-themselves/) - only 14% of Americans have actually tried ChatGPT personally.
I wonder how this bullshit-elimination effect will play out when LLMs truly become mainstream? And ironically, it's the push by Google, Microsoft, and other giants working to embed AI into every business product for the sake of productivity gains that is the most likely to get us there.
Agreed! I like this quote by scientist Roy Amara: "we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run." That's what I think will happen with these tools, too.
It's not much better than the LSTM we had in 2017, the only thing we have now is 64GB mobile-device, so in theory I can run a LLM on my phone now, but the text-gen is really not much better, I can compare I still have my live code side-by-side on my system running the same gpus the guy who did it all Karpathy is doing videos now explaining how the shit works, but the thing is he just has a bigger audience now but the tech is same-same
"Attention is the only thing you need" is six years old, there is nothing new under the sun.
Excellent article
Thanks Charles!!
I prefer to think of it as "bullshit banisher."
I like it! has a nice ring to it.
Also, can't remember if I mentioned this to you, but James Suzman's book "Work" is very good, and in it he talks about the term "bullshit jobs" explicitly. I really enjoyed it.
Will check it out, thanks Andrew
BULLSHIT for BULLSHIT-NATION a marriage made in heaven
https://bilbobitch.substack.com/p/chatgpt-one-ring-to-rule-them-all
Love this view.
Haven't thought of this aspect, very interesting. I suppose, as the ruling class sees free time as a danger, the use of ChatGPT to do pointless work will see as a danger too because it frees time for the employees. ChatGPT bans will ensue. Or the creation of additional tasks to compensate for the freed time resulting from the use of it.
This article should be read by anyone who is against/afraid of GPT.
Great article! It has many multifaceted dimensions. I think one more facet may be that so very many humans find AI threatening precisely because it can clearly show that what they do and what they’ve been up to for years is bullshit, and perhaps it always likely was. Yet even more damaging is the hugely decisive reveal linking their personal identities to bullshit, meaning precisely that the foundation of their identities are also bullish. That is a hard pill for a lot of people to swallow. It’s a real threat for people insecure in their core identities. Yet bullshit does pervade so much of human life.
One of the biggest bullshit jobs I ever saw was when I was an airline pilot flying in and out of China. As we would come through the entry doors of a major metropolitan airport there was one guy directing us to place our luggage in the metal detector while he scanned us with a wand. But that wasn’t the big bullshit job. After we passed through the wand guy and collected our luggage we had to stand in a line behind a tape. There was another guy to stop us at the tape and let the line build up. Then he would let us go. So, the essence of the job was to stop us. Then let us go. Stop us. Then let us go. Rinse and repeat. All -- day -- long. (I occasionally pondered how he would reply to his wife asking him if he had a good day at work that day - how would you reply? “Another great day honey. You should’ve seen me operate. Would you do me a favor? Just, kill me now”). In China these are called “rice bowl jobs”. They give people a sense of dignity where they get to wear uniform and do something invented which is just short of doing nothing at all. They get just enough of a wage to afford “rice bowl” of food. Such jobs are the epicenter of bullshit, but they give people a sense of dignity (supposedly) and keeps people from becoming disgruntled in an authoritarian regime building into an unmanageable mob. A lot of humans seem to crave a dignified ….. I guess you’d call it a ... Job. But I digress.
Another serious purpose I see for GPT-4, or ChatGPT, is to elevate true mediocrity to a more refined level. My 88 year old father recently delivered a reasonably good eulogy for his 94-year-old friend. It was above average by most standards, as he had been preparing it in his mind for years. When it came time to write his tribute for the local paper, I have to say it was pretty mediocre work. He’s an educated man with a Master’s degree in education, and I’m sure he thinks he’s an above average writer. But truth be told he’s a below average writer. This tribute was okay in that the sentiment he delivered was clearly sincere and genuine. But he just lacked certain skills to do a very great job. So, I volunteered to take his tribute and amp it up with GPT-4. I took my father’s contribution and applied the following prompt:
Please improve the writing of the following tribute: REMEMBERING DURAND BLANDING by Norm Toensing.
In with the marginal - out with the pretty damn impressive. The difference in quality was night and day. My father thought it was now a superb piece of writing and was incredibly appreciative. After proofreading it he only altered one word. My contribution took about five minutes. Yet in our small town I suspect my father’s reputation as an eloquent writer was secured. So, this tool also does have the ability to elevate mediocrity.