>I’ve been evangelizing the business benefits of enterprise technology for more than two decades, first as a marketing executive and more recently as a freelance writer. At IBM, I was known as a “Software Evangelist,” and my work always had me delivering secular sermons about the redemptive potential of the latest computer program or hardware.
>As an apostle of digital transformation, a prophet of workflow orchestration, I gushed with messages from the gods of tech about how products like AI and RPA would allow corporations to be born again—or at least generate higher earnings.
>And then, earlier this year, things started to go a little quiet. When you’re a freelancer, a slowdown is never good, but I grew more alarmed as a client notified me that they would no longer require my services because they were switching their content writing to AI. Others simply dropped me and didn’t send a note. Several more accused me of using AI in my writing, so why would they pay me?
This is a skill issue. If your writing is indistinguishable from AI, it has negligible value
> This is a skill issue. If your writing is indistinguishable from AI, it has negligible value
On the contrary, I think many such freelancers are way ahead of LLMs, but they are just too expensive in comparison. No amount of quality is worth the extra cost.
All this does beg the question of what happens when bots are talking to bots on either end.
I am imagining a world of long term context bots all negotiating with each other, over resources and services, in nanosecond time frames. In an environment where misinformation is both possible, and therefore competitively necessary.
The world hasn't been kind to evangelist-type people in the last few years. For a while, they got a whole lot of freedom, but the people I knew either got their leash pulled and now strictly need to shill company products or were let go/switched jobs.
I predict that the real market that's going to get absolutely decimated is the gig economy. The quality of AI may not be top notch but it's an easy drop-in replacement for one-offs that people previously would have farmed out to places like Fiverr (voice overs, logo, clip art design, copy editing, etc).
> I know graphic designers who used to outsource designs to Fiverr who are now just using Midjourney.
You know graphic designers who instead of designing would pay someone else to do it? That sounds like a cook who serves food bought from another restaurant or a carpenter who sells furniture bought from a store.
Maybe there’s detail missing from your explanation (perhaps you meant they outsource illustrations?), but if you’re paying someone else to do the core of your job can you still stay that’s your profession instead of being a middleman?
Graphic designers don't often don't create every little thing that goes into what they do, and often don't do their own illustrations (there are exceptions, but this is usually done by dedicated illustrators).
Exact example I was thinking of: graphic designer does social media post for BigCorp, let's say Christmas themed.
They're going to get an illustration from somewhere, do a bit of photoshop, add some text over the top, and then post it to social media (oversimplified, I know).
Especially for a small shop where they don't have full time illustrators a lot of the illustration work gets outsourced. These days, some of it gets outsourced to Midjourney.
I like reading random blogs and articles about tv shows I follow. This week, more than one of this articles I found were AI generated. Pretty obvious, bad and formulaic content. Do I need to flag this websites from my end? Do I need to whitelist my sources? It's a shame.
You need to evolve with changes in tech. A technology evangelist should focus more on networking, attending conferences and talking to executives, in person if possible. That is, focus on the human touch.
Another avenue is that of influencer on social media.
If we was mainly doing writing articles, then that made him easy to replace.
Learning new things takes time. When we were inventing new machines once per generation, that means workers needed to learn new skills once per generation, which is completely achievable.
If we can train AIs faster than human beings can learn new things, there's no way for people to keep up. Even if AIs are a bit slower that's still an issue. Imagine getting a new masters degree every 5 years.
Why is that an automatically agreeable opinion? There are a lot of people in professional trades who go to school once in their career, and then periodically augment their skills. But they don't go and get a whole new degree every few years just to continue to have a job
> How do you think the AI software gins up writing that sounds like something I wrote? It does so by stealing.
If I write a text in the style of, say, Vonnegut, is that stealing? Why is it stealing when an AI does it?
AIs don't steal any more than a human who has read someone's writings is stealing when writing. Whether we want AIs or not is a completely different discussion, and a valid one, but "we don't want AIs, therefore we'll claim they're stealing" is not valid, in my opinion.
That's an incorrect analogy. A more proper one would be if you stole Vonnegut's books, then wrote your work in his style after learning it from those books.
This is really a nitpick. If I can write books in the style of Vonnegut and he loses his entire income (who knows how many millions of dollars), it doesn't matter if I first bought some of his books for $100 or just stole them.
For AI, they pulled massive amounts of data off the internet. For any copyrighted material, I would consider this fair use, seeing as they used trillions of words of training, any individual work is a tiny fraction. Plus if you ask ChatGPT to give you the text of Cats Cradle, it won't do it, so there is no direct copying happening. Maybe the analogy is walking into a bookstore, reading his books, and then leaving without paying. But it doesn't really matter, because they can retrain without any of that without affecting the model quality all the much, the real magic is transformers + RLHF.
> A.I. companies have acknowledged in research papers that they rely on shadow libraries. OpenAI's GPT-1 was trained on BookCorpus, which has over 7,000 unpublished titles scraped from the self-publishing platform Smashwords. To train GPT-3, OpenAI said that about 16 percent of the data it used came from two "internet-based books corpora" that it called "Books1" and "Books2." According to a lawsuit by the comedian Sarah Silverman and two other authors against OpenAI, Books2 is most likely a "flagrantly illegal" shadow library.
If “be ok” means being comfortable with not having the income to them from people paying them to write that stuff, then no, I don’t think the artist cares whether the AI pirated or paid the $500 to buy all 50 of their books (or whatever).
If “be ok” means that legally they are limited because someone bought their book and chose to use it by feeding into a training algorithm, then yes. If I steal a book from your house, it doesn’t matter if I stole to read myself or to feed into an algorithm.
But if I buy a book, or borrow it from a library, then it’s legal for me to have a computer scan it and feed into a training program. I just can’t distribute copies or otherwise infringe on the copyright.
I still argue that it is not stealing what they do.
People always learned from others and just because ai can do it much better doesn't change this.
And yes the 'why would I make myself obsolete ' is some type of fallacy.
Your skills should not be easily exchangeable otherwise you don't have a skill relevant to the might god of capitalism. That's how capitalism optimizes.
You either up skill or start pushing for a global income.
He mentioned Upwork. Personally, I feel like Upwork has a near monopoly on easy-to-access freelance work at this point. Except for jobs that you find from direct acquaintances or friends or through fairly long involved searches. And I am probably more isolated than most, but I suspect that many people's networks get saturated pretty quickly with job requests when things are tight.
I rely on Upwork when I run low on time to find the next gig quickly between work on my own startup.
I believe that within a few years they may start going a little along the path of Stack Overflow. They will have a whole section of AI "staff" that are literally competing with the real people on the site for many types of jobs. This will enrage the current users even more than what has happened with Stack Overflow.
But I don't think they really have a choice. They may already be working on it.
If Upwork doesn't do it then there will likely be another site along those lines specializing in "AI employees".
Yes, it's extremely challenging for that reason. If you have another option please let me know. I just don't really have a network of people to ask and generally don't have time or know how to build one. So I would be happy for another site like Upwork without so much competition.
Although I think if you are willing to work for relatively low rates like me, being in the US makes you stand out and possibly easier to get hired.
Upwork does have an option to post jobs that only allow US contractors. And you can search for only those types of jobs if you are in the US. But often I take that off because it restricts the search quite a bit.
I think the reality is that since more and more people are working remotely, it's a global talent pool and kind of a race to the bottom in terms of rates.
I just wish that there was more competition in terms of sites like that.
There are certain jobs that are just always bound to be automated (any work near a bandwagon or a phone), instead I think the bigger issue is how our economic model is so reliant on maximum acceleration of adoption that we never stop and ask if people are fine with the jobs they have, for instance there's a very well known phenomenon in the military where a lot of soldiers don't really wants to climb the ranks because it means it stops being a field job and instead turns into a desk job.
Its funny how 10 years ago the luddities were usually seen as out of touch or just greedy dumb while the CEOs was seen as pioneers.
And now slowly we see people sympathize with the luddities.
What I wonder is will this automation cause yet another labor movement? If so how different will it be?
> I think the bigger issue is how our economic model is so reliant on maximum acceleration of adoption that we never stop and ask if people are fine with the jobs they have
This was never and wont ever be the point of a job though. Its what the market demands, not what people want out of a job.
And I agree, but historically this was not the case and it seems that at some point we either embrace that (very unlikely) or we make the transitioning period the smoothest we can.
Again the last time this happen we got the labor movement in protest of how terrible the transitioning was.
> After offering a legally-binding attestation that I have not, and never will use AI in my writing
That's probably an ill-advised attestation to give from someone deeply familiar wtih AI. All of the grammar tips that Word and other English language editors give you, or context-aware spell-checking? Those are already (or soon will be in the worst case) using AI.
To the extent that your final word choice is influenced by these AI agents, you're using AI in your writing.
Not to mention that predictive text on mobile keyboards is also AI. SwiftKey used a neural network to suggest the next top-k words a decade ago and now all mobile keyboards do it
> All of the grammar tips that Word and other English language editors give you, or context-aware spell-checking? Those are already (or soon will be in the worst case) using AI.
Will you also count any ML algo that recommends some book, podcast or video which affects what the author writes? How far back do you want to go?
In practice "using AI in writing" means a simple thing: "you give a prompt and get text written for you".
> It’s a Kafkaesque experience to try to prove that you have not stolen writing from a machine that stole your writing in the first place. ... My business isn’t dead, but it’s wounded
Something similar I thought[0] would be happening.
Not only commercial LLM operators abuse copyright, take away profit and turn every author into a ghost writer for their products—but they also make your (now former) audience not believe you have done original work that was expropriated from you, usually without your consent.
In case of Hugh Taylor this meant tarnished reputation. Over longer term, it means elimination of incentives to share information openly, as you stand nothing to gain in terms of respect if the next person automatically assumes that you have simply paid for some LLM.
>> but they also make your (now former) audience not believe you have done original work
I've experienced this.-
PS. One of the (most) worrisome aspects of AI might just be this: The general "depreciation" of writing, as with the "depreciation" of "truth" in this "post-truth" age, when absolutely everything can be assumed to have non-human origin.-
48 comments
[ 2.7 ms ] story [ 392 ms ] thread>As an apostle of digital transformation, a prophet of workflow orchestration, I gushed with messages from the gods of tech about how products like AI and RPA would allow corporations to be born again—or at least generate higher earnings.
>And then, earlier this year, things started to go a little quiet. When you’re a freelancer, a slowdown is never good, but I grew more alarmed as a client notified me that they would no longer require my services because they were switching their content writing to AI. Others simply dropped me and didn’t send a note. Several more accused me of using AI in my writing, so why would they pay me?
This is a skill issue. If your writing is indistinguishable from AI, it has negligible value
On the contrary, I think many such freelancers are way ahead of LLMs, but they are just too expensive in comparison. No amount of quality is worth the extra cost.
All this does beg the question of what happens when bots are talking to bots on either end.
What dynamics and sub-dynamics would arise?
An hour goes by, like a century for us.
A strange dystopia.
http://www.antipope.org/charlie/blog-static/fiction/accelera...
Yes, now. But the point is, that wasn't the case before.
Also, large part of what used to be done with stock images. No need to search for the perfect stock image when you can just create it.
You know graphic designers who instead of designing would pay someone else to do it? That sounds like a cook who serves food bought from another restaurant or a carpenter who sells furniture bought from a store.
Maybe there’s detail missing from your explanation (perhaps you meant they outsource illustrations?), but if you’re paying someone else to do the core of your job can you still stay that’s your profession instead of being a middleman?
Exact example I was thinking of: graphic designer does social media post for BigCorp, let's say Christmas themed. They're going to get an illustration from somewhere, do a bit of photoshop, add some text over the top, and then post it to social media (oversimplified, I know).
Especially for a small shop where they don't have full time illustrators a lot of the illustration work gets outsourced. These days, some of it gets outsourced to Midjourney.
Another avenue is that of influencer on social media.
If we was mainly doing writing articles, then that made him easy to replace.
If we can train AIs faster than human beings can learn new things, there's no way for people to keep up. Even if AIs are a bit slower that's still an issue. Imagine getting a new masters degree every 5 years.
At some point, we're all pretending to work 100% of the time, and can just... stop.
If I write a text in the style of, say, Vonnegut, is that stealing? Why is it stealing when an AI does it?
AIs don't steal any more than a human who has read someone's writings is stealing when writing. Whether we want AIs or not is a completely different discussion, and a valid one, but "we don't want AIs, therefore we'll claim they're stealing" is not valid, in my opinion.
Vonnegut gets no money from you learning, but nothing bad or illegal happened.
But they didn't do that
> Or checked them out from the library.
But they didn't do that either.
For AI, they pulled massive amounts of data off the internet. For any copyrighted material, I would consider this fair use, seeing as they used trillions of words of training, any individual work is a tiny fraction. Plus if you ask ChatGPT to give you the text of Cats Cradle, it won't do it, so there is no direct copying happening. Maybe the analogy is walking into a bookstore, reading his books, and then leaving without paying. But it doesn't really matter, because they can retrain without any of that without affecting the model quality all the much, the real magic is transformers + RLHF.
I suspect they definitely did as this is just an analogy we invented in this thread.
And in real life, I’m guessing that people training LLMs are definitely buying ebooks and not pirating them.
https://www.nytimes.com/2023/07/22/business/dealbook/extreme...
Not to mention that the article's author called it "stealing" even though the articles in question are free to read.
If “be ok” means that legally they are limited because someone bought their book and chose to use it by feeding into a training algorithm, then yes. If I steal a book from your house, it doesn’t matter if I stole to read myself or to feed into an algorithm.
But if I buy a book, or borrow it from a library, then it’s legal for me to have a computer scan it and feed into a training program. I just can’t distribute copies or otherwise infringe on the copyright.
People always learned from others and just because ai can do it much better doesn't change this.
And yes the 'why would I make myself obsolete ' is some type of fallacy.
Your skills should not be easily exchangeable otherwise you don't have a skill relevant to the might god of capitalism. That's how capitalism optimizes.
You either up skill or start pushing for a global income.
I rely on Upwork when I run low on time to find the next gig quickly between work on my own startup.
I believe that within a few years they may start going a little along the path of Stack Overflow. They will have a whole section of AI "staff" that are literally competing with the real people on the site for many types of jobs. This will enrage the current users even more than what has happened with Stack Overflow.
But I don't think they really have a choice. They may already be working on it.
If Upwork doesn't do it then there will likely be another site along those lines specializing in "AI employees".
Although I think if you are willing to work for relatively low rates like me, being in the US makes you stand out and possibly easier to get hired.
Upwork does have an option to post jobs that only allow US contractors. And you can search for only those types of jobs if you are in the US. But often I take that off because it restricts the search quite a bit.
I think the reality is that since more and more people are working remotely, it's a global talent pool and kind of a race to the bottom in terms of rates.
I just wish that there was more competition in terms of sites like that.
Its funny how 10 years ago the luddities were usually seen as out of touch or just greedy dumb while the CEOs was seen as pioneers.
And now slowly we see people sympathize with the luddities.
What I wonder is will this automation cause yet another labor movement? If so how different will it be?
This was never and wont ever be the point of a job though. Its what the market demands, not what people want out of a job.
Again the last time this happen we got the labor movement in protest of how terrible the transitioning was.
That's probably an ill-advised attestation to give from someone deeply familiar wtih AI. All of the grammar tips that Word and other English language editors give you, or context-aware spell-checking? Those are already (or soon will be in the worst case) using AI.
To the extent that your final word choice is influenced by these AI agents, you're using AI in your writing.
> All of the grammar tips that Word and other English language editors give you, or context-aware spell-checking? Those are already (or soon will be in the worst case) using AI.
Will you also count any ML algo that recommends some book, podcast or video which affects what the author writes? How far back do you want to go?
In practice "using AI in writing" means a simple thing: "you give a prompt and get text written for you".
Something similar I thought[0] would be happening.
Not only commercial LLM operators abuse copyright, take away profit and turn every author into a ghost writer for their products—but they also make your (now former) audience not believe you have done original work that was expropriated from you, usually without your consent.
In case of Hugh Taylor this meant tarnished reputation. Over longer term, it means elimination of incentives to share information openly, as you stand nothing to gain in terms of respect if the next person automatically assumes that you have simply paid for some LLM.
[0] https://news.ycombinator.com/item?id=34665676
I've experienced this.- PS. One of the (most) worrisome aspects of AI might just be this: The general "depreciation" of writing, as with the "depreciation" of "truth" in this "post-truth" age, when absolutely everything can be assumed to have non-human origin.-