I’m sorry Dave, I can’t do that.
I don’t really think that AI is such a big risk as all that. Although it may end up being a risk to the income of the Washington Post’s writers.
This is it. I said a while ago that there was going to be hysteria over AI, because this time an advance in tech isn’t coming for assembly line workers or farmers, it’s coming for the chattering classes, and they have the megaphone.
So if jobs are lost in manufacturing, they can just ‘learn to code’. But if they are lost in the media and academia, well, the world just might be collapsing!
I think it’s okay to question some of this. Maybe AI can write your term paper, but I don’t think it should be. Any more than outsourcing your term paper to a smarter student. That’s not the point.
Automating physical labor away is one thing. I don’t want to work in a coal mine either. But speaking and writing are higher functions of what it means to be human and the potential for far greater atomization and isolation and meaninglessness of the human experience in this realm is significant. I already have substantial concerns about the impact of the Internet and isolation of Covid and this is only going to make those elements worse.
Wonder if he’ll use ChatGPT for that.
What an idiot. The AI works great if you know what to use it for. You can’t just dive right in and assume it’s good for legal briefs.
That’s a false and cynically anti-intellectual spin on what the real concerns are. The replacement of manual labour with machines, AI or otherwise, is generally a long-term net benefit for everyone. It improves productivity and allows humans to take on more dignified and cerebral roles. The fear now is that AI is intruding on higher levels of intellectual function and on the realm of creative expression itself, the last frontier of humanity. I believe these fears are very much misplaced, and AI just brings us a new set of more sophisticated tools that improve our lives, even if, inevitably, it does displace some jobs.
I am cynical about this.
Who says humans in general want ‘cerebrtal roles’? My grandfather farmed all his life, and loved it. The idea that he should take on more ‘cerebral roles’ would have been as anathema to him as telling a lawyer he can learn to farm.
In any event, there are huge productivity gains to be had by AI. We haven’t even begun to see the true impact of this tech on the armies of people who deal in information. Backoffices will probably be able to run with a fraction of the people they use now. How many government worker bees can be eliminated? How many people out there writing reports, ad copy, summaries, prospectuses, briefs, powerpoint presentations, and other mundane communications will be needed in the future?
The fact that a business can be run by a fraction of the staff + AI is very good news for equality. The cost of starting up and running a small business just plummeted, as did the difficulty in learning how to do it. No longer will people have to go hat in hand to venture capitalists and other money people to capitalize on an idea.
This is the best thing to happen to average people in a very long time. They are no longer required to hire expensive, wealthier people to finance them or provide their services. You can probably run a million dollar business as a sole proprietor with the help of suitable AI assets. That’s a great thing for equality. Even greater, the tech is accessible to just about anyone, since you ‘program’ it in natural language.
And there is a lot of tedium in many office jobs that are just as demeaning as ‘manual labor’. Being a paper shuffler is not fun.
But what did he love about it? Was it the country living, fresh air, wide open spaces, and the satisfaction and income of a good harvest? Or was it getting up at 5 AM for a long day of tedious and often back-breaking labour in the hot sun, day after day?
Increasing automation including AI isn’t going to displace farmers. It’s going to make their lives easier. The rest of your post generally seems to support my points.
I include that sort of work in my definition of “manual labour”. Perhaps I should have said “menial labour”.
Working with his hands, being his own boss, getting to see the fruits of his labor grow and mature, being outdoors, being self-sufficient working with machinery and animals, etc.
This was a man who would be driving us home from church and see a nice wheat field, and he’d have to stop and wade into it and examine it and proclaim its strengths and weaknesses. He loved everything about farming. Even the long hours and the back-breaking work.
It’s actually office work that is abnormal for people. I’m the only one of my family who could even tolerate it. My grandfather’s idea of hell would have been plonking him behind a desk and making him write reports or something.
Anyway, Mechanization came for thr farmers lalready. Automation came for the assembly line workers, telephone operators, etc. Now AI is coming for some of the office workers.
What the hand-wringers aren’t getting is that these changes improve productivity, which grows the economy and creates jobs. Reducing the friction of starting a business and capitalizing on an idea is both egalitarian and increases economic growth and the rate of innovation. Lowering the dead-weight management costs of corporations will make them more efficient.
Lots of white collar people will lose jobs, and lots more jobs will be created. This has been going on since the Luddites tried to protect their weaving jobs. It’s responsible for most of the wealth we have today.
As I said, I think the response this time will be different, because the people affected are richer and have a big megaphone. Their problems will be deemed more critical than those of farmers or assembly line workers or loom operators.
I think it’s misleading to refer to what an AI does as “hallucinations”. When the AI came up with fake cites for that lawyer, it wasn’t malfunctioning and doing something wrong. It was doing the same thing it always does: It was bullshitting. Often, the bullshitting AIs are factually correct, and sometimes, they’re factually incorrect, but the distinction is irrelevant to a bullshitter. Bullshit is bullshit, regardless of whether it’s true or not.
And while AIs might eventually take over any given job, I think that the only jobs that the current crop of AIs (ChatGPT and its ilk) will take over are the professional bullshitters. Of which there are plenty; there seems to be an inherent human demand for a certain level of bullshit. But I find it hard to mourn professional bullshitters being outcompeted out of work.
And office assistants, and spreadsheet jockeys, quants, accountants, purchasing people, ISO managers, presentation people, tech writers, ad copy writers, corporate communications departments, writers of technical reports, QAs and programmers, database analysts, business analysts, content moderators, web developers, business graphic artists, UI developers…
Much office work comes down to communication. Employees send out emails through a partially automated process. Bosses of these employees know what is in the emails but don’t know how to send them out. The employees know how to send them out but don’t know how to automate the process OR DESCRIBE THE PROCESS LOGICALLY. IT lacks resources and doesn’t understand the business. Typically a third party is needed between IT and the people sending the emails out, either a business analyst or a programmer with business experience.
If it were simply a matter of providing the correct prompts to an AI the process would already be much simplified.
Imagine an AI that reads every single email that passes through the company’s mail server, and knows when someone missed an important email or forgot to reply or forgot to send their draft.
These sorts of things have already been automated mostly.
Also keep in mind that many people aren’t going to want an AI that knows everything the company is doing. For many reasons. Including regulatory. Who wants an AI who will blab away to auditors? You’re going to find that this sort of info is going to move out of reach of any AI.
One may hope. There is a very interesting quote in a rather lengthy discussion by Steven Wolfram.
“So how is it, then, that something like ChatGPT can get as far as it does with language? The basic answer, I think, is that language is at a fundamental level somehow simpler than it seems. And this means that ChatGPT—even with its ultimately straightforward neural net structure—is successfully able to “capture the essence” of human language”
Full article here : What Is ChatGPT Doing … and Why Does It Work?—Stephen Wolfram Writings
One wonders how this might tie in with Chomsky’s ‘generative grammar’ ideas…though of course Chomsky hasn’t really been working as a linguist for a long time now…?
And by the way, I do not want to see anything resembling the current crop of chatbots entrusted with any mission-critical or life-threatening decisions. We have seen quite enough examples of how they generate complete bullshit (as Chronos puts it). Unfortunately it will probably happen as a cost-cutting exercise.
I doubt that. By your logic, companies would still be keeping their books in physical pen and paper format, because that’s much harder to trace than digitized copies; further, even if companies DID choose to digitize, they’d NEVER agree to move sensitive data from their local server to the Cloud, as they wouldn’t give up control of the data.
Fact is, digitizing all of your data provides enormous benefits. So does moving everything to the Cloud. So companies are making these choices (pretty much universally for digitizing data, and more and more for Cloud solutions), and I guarantee they will do the same for AI once it is profitable to do so.
Glad to know that our lawyers are wrong and the records we don’t retain will now be retained because AI. /s
ChatGPT’s “blabbing” would be nearly useless to auditors or other enforcers, because there’s no way to tell how accurate it is.
In the Intelligence, Literacy, Carpet thread, people have been testing ChatGPT on questions that, combined, only 4% of people get right.
The first was a Q about calculating the cost of carpet sold by the square yard, to cover a room whose dimensions were given in feet.
ChatGPT got the answer right. But I wondered if this might be because the test used or something close to it cropped up in training data. So I gave it a similar test.
It failed to get the right answer, but in quite interesting ways.
Prompt:
Suppose you want to sow a grass field which is 1503 feet by 2184 feet, and you purchase grass seed at 30% off in a sale. It will cost £7.89 per square yard of coverage. How much will it cost to sow the field?
Response:
To calculate the cost of sowing the grass field, you’ll need to follow these steps:
- Calculate the area of the field in square feet: Area = Length x Width Area = 1503 feet x 2184 feet Area = 3,281,352 square feet
- Convert the area to square yards since the grass seed price is given per square yard: 1 square yard = 9 square feet Area in square yards = 3,281,352 square feet / 9 square feet/square yard Area in square yards = 364,594.67 square yards (approximately)
- Calculate the regular price of the grass seed for the entire field: Regular price = Area in square yards x Price per square yard Regular price = 364,594.67 square yards x £7.89/square yard
Regular price = £2,877,873.66 (approximately)
- Apply the 30% discount to find the sale price: Sale price = Regular price - (30% of Regular price) Sale price = £2,877,873.66 - (0.30 x £2,877,873.66)
Sale price ≈ £2,014,511.56
So, it will cost approximately £2,014,511.56 to sow the field with grass seed at a 30% discount.