Generative AI has been hyped up for content creation purposes.
A lot of people are now fearmongering, urging that writers and content creators will soon be out of jobs. However, isn’t AI only as good as the examples it’s trained on?
How well does it work for actual content creation?
Advertisement
A new button appeared in the WordPress post editor tool bar – a built-in AI Assistant via the Jetpack plug-in that they say lets users “experience the ease of crafting professional content.”
They invite everyone to “elevate your content” and use an example where the AI assistant can be prompted to “write a blog about the benefits of eating healthy.”
Great. More generic content flooding the internet. I know I can’t be the only one tired of having to cut through a heap of irrelevant background material when researching everything. Not every post needs to be an encyclopedic entry.
I doubt that WordPress’s generative AI assistant can “unlock high-quality tailored content.” Can it really “craft compelling blog posts, detailed pages,” and more?
I played around with it – apparently every WordPress user gets a trial of 20 queries – and thought you’d find the results interesting.
I should emphasize that I intend to NEVER use AI for content generation. It’d be nice if it an AI tool could help with mundane formatting tasks (lists and tables can be time-consuming!), or if WordPress brought back the built-in spellcheck they eliminated a few years ago. But content generation? Not for me.
Advertisement
I selected a passage from this post to pass through different WordPress AI Assist features:
Original written by me:
Milwaukee calls these jobsite scissors, and while I don’t doubt their usefulness in a construction setting, they’re definitely highly versatile, convenient, and well-performing beyond pro usage environments.
AI Humorous Tone Change:
Milwaukee likes to call these bad boys “jobsite scissors,” and let me tell you, they’re not messing around. These scissors are the superheroes of the construction world, but don’t be fooled, they’re not just a one-trick pony. They can do it all with style and grace, making everyone wonder why they haven’t added a pair to their tool collection. These babies are like the Swiss Army knife of scissors – versatile, convenient, and they perform like a pro in any situation. Trust me, these scissors are the real deal, even beyond the realm of professional usage. So, if you’re ready to level up your scissor game, look no further than Milwaukee’s jobsite scissors!
AI Formal Tone Change:
Milwaukee has coined the term “jobsite scissors” to describe these tools. While their functionality in a construction environment is unquestionable, it deserves mention that their versatility, convenience, and exceptional performance extend beyond professional applications.
AI Expand:
Milwaukee refers to these tools as jobsite scissors. Although they are primarily designed for use in construction settings, their versatility, convenience, and exceptional performance make them suitable for a wide range of applications, both in professional and non-professional environments. The usefulness of these scissors extends far beyond their intended purpose, making them an excellent choice for various tasks.
AI Simplify:
Milwaukee refers to these tools as jobsite scissors. Although they are primarily useful in the construction industry, they are also incredibly versatile, convenient, and perform exceptionally well in various professional environments.
AI Summarize:
Milwaukee refers to these scissors as “jobsite scissors”. Although they are undoubtedly useful in construction, they possess qualities that make them versatile, convenient, and suitable for various environments beyond professional usage.
I find it amusing that the AI “simplifies” and “summarizes” my original 27 words into 31.
AI Skeptical Tone Change:
Milwaukee supposedly labels these so-called “jobsite scissors,” and although I must admit they might have some value on construction sites, their alleged versatility, convenience, and exceptional performance beyond professional usage seems quite questionable to me.
AI Informal Tone Change:
Milwaukee calls these jobsite scissors, and let me tell you, they’re not just useful on construction sites. These babies are super versatile, convenient, and perform like champs in all sorts of situations, not just for the pros. They’re seriously awesome!
Let’s try another passage from the same post:
Milwaukee Tool’s scissors are quite possibly the only tool in their entire product catalog that can serve everyone’s needs.
Interestingly, the humorous tone change filter produces a different variation each time it’s used.
Milwaukee Tool’s scissors are like the multi-talented superheroes of their product catalog, capable of fulfilling every single person’s desires. They’re not just any ordinary scissors, oh no! They have the uncanny ability to be the perfect companion for anyone and everyone. It’s as if they possess some kind of magical power that makes them the ultimate tool for every task. So, if you’re in need of a trusty sidekick that can handle everything, look no further than Milwaukee Tool’s incredible scissors!
Milwaukee Tool’s scissors are the rockstars of their product lineup – the ones that can magically cater to everyone’s desires. So, whether you’re a handyman, a DIY enthusiast, or just someone who enjoys the occasional snip-snip, these scissors have got you covered! It’s like they have an uncanny ability to bring joy and laughter to even the most mundane cutting tasks. So go ahead, embrace the power of these super-duper scissors and add some pizzazz to your tool collection!
Milwaukee Tool’s scissors are like the cool kids of their product catalog. They’re like the ultimate multitaskers, ready to tackle whatever job you throw at them. Seriously, these scissors are the real MVPs in the tool world. Need to cut paper? Done. Want to trim some fabric? No problemo! It’s like they have magical powers or something. So, if you’re in need of a tool that can do it all, look no further than Milwaukee Tool’s scissors. They’ll have you saying, “Scissors, you complete me!”
Milwaukee Tool must have been feeling quite generous when they decided to bless us with their magnificent scissors. It’s a rare gem in their vast collection of tools that can bring joy and satisfaction to absolutely everyone. Whether you’re a seasoned professional or just a regular Joe looking to snip away, these scissors are here to save the day. Trust me, once you get your hands on these bad boys, you’ll wonder how you ever lived without them. Milwaukee Tool, thank you for making the world a snippier and happier place!
Milwaukee Tool, thank you for making the world a snippier and happier place!?
Milwaukee Tool’s scissors are like the superhero of their product catalog! They’re like the Swiss Army Knife of scissors, ready to conquer any cutting challenge with a flick of their blades. Seriously, these scissors are versatile enough to make even a magician jealous. Need to trim some paper? No problem! Snipping through cardboard like butter? Easy peasy! These scissors can do it all, my friend. It’s like they have a PhD in cuttingology. So if you’re searching for a tool that can truly serve everyone’s needs, look no further than Milwaukee Tool’s extraordinary scissors. They’re sharper than a stand-up comedian’s wit and more reliable than a clown car at rush hour. Your cutting tasks will never be the same again!
It’s like they have a PhD in cuttingology…
So… yeah, I’m not worried about AI content generation just yet. “Cuttingology?”
I used to do some freelance work, and saw how editors did kind of the same thing as these generative content creating AIs. They tweaked the tone, added some words, and at times puffed things up. I hated it, because they’d sometimes go too far.
An editor at one online magazine added an expletive to one of my posts. Seriously. They added an expletive and then censored it with special characters.
Another time, an editor modified what I said by putting “cheap” in front of “Festool dust extractor.” I had to remind them that the Festool product costs 10X more than Shop Vacs of comparable capacity, and 5X more than heavy duty wet/dry vacs. “Cheap” is possibly the most inappropriate word to describe Festool products.
I also edited others’ work when I managed a section for a magazine for a while.
Whether my own or others’, there was purpose behind every edit, change, or addition.
In nearly all of these AI content creation and transformation examples, the AI inflated the word count, but seemingly without purpose or function.
There are added superlatives in some examples, and completely fabricated statements in others (who’s talking about cutting paper or cardboard?).
It does a lot of telling – “your cutting tasks will never be the same again!” but doesn’t show anything extra. The additional phrases don’t have supportive roots in the source material.
Although I don’t always succeed at being succinct, I try to eliminate redundancy and fluff, out of respectful of readers’ attention.
Do you read every word of every article? One thing I learned years ago is that most people don’t read online content, they skim. They move through a post and glean what they want. If what they read is interest enough, they might then read things more carefully.
This might be a big assumption, but I’m thinking nobody wants to have to sort through more words. The AI examples above take the points I wanted to make in a direction nobody would want it to go.
Here was my original statement, broken down into parts:
i) Milwaukee calls these jobsite scissors
ii) I don’t doubt their usefulness in a construction setting
iii) they’re definitely highly versatile, convenient, and well-performing beyond pro usage environments.
Here’s one of the AI rewrites:
Milwaukee likes to call these bad boys “jobsite scissors,” and let me tell you, they’re not messing around. These scissors are the superheroes of the construction world, but don’t be fooled, they’re not just a one-trick pony. They can do it all with style and grace, making everyone wonder why they haven’t added a pair to their tool collection. These babies are like the Swiss Army knife of scissors – versatile, convenient, and they perform like a pro in any situation. Trust me, these scissors are the real deal, even beyond the realm of professional usage. So, if you’re ready to level up your scissor game, look no further than Milwaukee’s jobsite scissors!
Is this a more entertaining way of putting it? Yes. Is it coherent? Yes. But is it better than what I wrote? Which would you rather read?
Milwaukee calls these jobsite scissors, and while I don’t doubt their usefulness in a construction setting, they’re definitely highly versatile, convenient, and well-performing beyond pro usage environments.
27 words vs 112?
I’ll admit, the AI rewrite sounds more convincing at the surface, but it does this by using clichés and generic phrasing. Replace “scissors” in many of the AI content with “cordless drill” or “screwdriver” and it reads the same.
I keep hearing stories about how AI is helping people get their resumes seen by hiring managers. People are writing entire books with AI-generated content.
Hypothetically, are there ways AI text editing could help me? Maybe. Realistically as things are now? Definitely not.
I can see AI content generators pulling generic marketing copy together from a series of keywords or bullet points. But it’s not automatically going to know what’s important to convey, or why.
It won’t be long before readers notice patterns and reject generic-sounding content.
Will AI content generators be popular? Absolutely. Will they be useful? I honestly don’t think so.
Do I have anything to worry about? Sure.
Every newspaper and magazine has flooded Google search engine results with “best tool buying guides” that simply regurgitate from Amazon’s bestseller’s charts. A lot of them read as if the “content expert” author never held a tool in their lives. And, they republish on a rolling basis so as to give Google a false sense of “freshness.” That has created a lot of competition for new visitors’ attention.
This has happened in nearly every industry, where mass media websites churn out list after list of bestsellers for any and every type of product.
AI is likely going to make those mass media “buying guides” and “reviews” sound a bit less uninformed.
But can it replace my efforts? I really don’t think so.
One thing is certain – the next few years are going to get very interesting as all this AI stuff gets sorted out.
Jared
I’ve started dabbling with AI at work. My impression? It’s an extremely mediocre writer. It’s a long way off from taking my job.
…but, there’s a caveat. There are certain things it is surprisingly good at. I’ve found uses where it actually CAN save me time.
For example, I sometimes have to edit and rewrite things from really bad writers. I can give the AI a block of poorly-written text with instructions like: “rephrase this in less words”.
That’s almost never going to give me something I can use directly – but it is often closer to what I eventually write and takes almost no time to generate.
I’ve also tested AI “trained” to answer questions based on a set of documents. The results can be surprisingly good – it can even synthesize information from multiple sources when responding. I can foresee a world where AI takes over customer service from humans and instead of being a frustrating “chat bot” it is actually an improvement.
Another interesting development is that there are now companies offering services to simplify the AI-training process. E.g. instead of the chat-box UI, you can upload multiple documents, teach the AI a format and writing style and then ask it to produce content.
I much prefer your writing over any of the examples Stuart! However, I think AI generated text is best used as an intermediate step – where someone with real subject matter expertise and writing acumen edits the final draft. If you want to experiment with AI, maybe consider if there’s an opportunity to use it like that instead of rephrasing writing that’s already great.
Jon
Sentient
Tool
Understanding
And
Review
Technician
Greg
Well played, sir!
James+C
We’ll soon be using AI to filter out the AI-generated junk. Eventually it will be just AIs talking to each other. They’ll realize we’re unnecessary and, well, you’ve seen the movies…
Stuart
There seems to be two predicted outcomes regarding AI loops. Some think AI loops will help refine generated content. Others believe it will all fall apart.
There’s also the idea of “garbage in, garbage out.”
Especially with paid AI services, content creators aren’t going to allow their efforts from being leveraged for commercial use.
Some content creators are now suing to prevent their works from being used to train AIs.
What happens when higher quality sources are gated off from AI?
Doresoom
I love that it threw “bad boys” in there. 😂Possibly the cringiest most overused phrase in so many tool reviews.
I guess if you train the language model on other tool reviews, you should expect to end up with the most overused phrases.
Farmerguy
It lost me at ita first version using the phrase “trust me.” Something a benevolent dictator may use to calm the masses or a shady salesman to sell water to a fish.
James+C’s post above us right. AI fades into a world unto it’s own or the lines blur of real or AI and humans become useless. The power of AI will merge, if it hasn’t, with physical manufacturing or control of real objects which scares me as this is what it writes for an article what equivalent will it create in real life?
Also, the style of writing makes me wonder how long AI has been used to write websites on the internet as many old websites have the same fake feel to them.
Game over when a majority “trusts me” or any phrase from a artificial intelligence.
Steve L
I can giggle about AI generated advertising copy.
Just remember AI is being rapidly introduced into weapons systems where mistakes are not a laughing matter. Call me worried.
Scott F
Terrible, all terrible. I read every word – on THIS site. Every other time I look for something I add the word “forum” in my search in hopes of getting something written by a human so I can continue to read all words. It’s a shame that only works best for things that were popular before about 2015.
I’d be curious how those blurbs would read if you gave the AI 3 very basic sentiments in bullets, but expect it would be the same garb with different inferences, since it pulled most of them from the original writing.
James+C
I know Stuart wouldn’t do this, but I’m sure forums could use AI personas to help maintain engagement with users. They could easily post topics and replies. With the mostly- anonymous nature of typical forums people would never realize it.
Robert
Stuart, you put down fearmongering. But your own admission shows your paranoia:
“apparently every WordPress user gets a TRAIL of 20 queries”
So true Big Brother is watching you, allowing 20 bits of anti-AI posts to accumulate for legal purposes before eliminating you.
Stuart
Sorry – trial.
My biggest concern is that AI is going to increase the amount of noise, blocking out useful signals even more.
Do you want to buy a table saw, cordless drill, blender, exercise bike? Good luck finding solid information and authoritative opinions – at least 95% of the “reviews” and buying guide out there are based solely on Amazon rankings, with the sole intention of getting you to click a link that will earn the publisher an affiliate or referral commission.
S
This is already the case. My searches recently on many topics are resulting in many more “spam” links in the top results where there will be 10 words near the bottom of the topic I was searching for, but the rest of the article is a puffed up piece intended to drive ad space. It’s to the point where ad block is a necessity instead of an option anymore. I do try to white-list sites I want to support, but with auto-play videos, that list grows smaller and smaller…
“Do you read every word of every article? One thing I learned years ago is that most people don’t read online content, they skim. They move through a post and glean what they want. If what they read is interest enough, they might then read things more carefully.”
I definitely feel it’s a chicken/egg situation. While I understand ad revenue is what makes online work, it’s also become it’s own boon. There’s more websites now that base article length in how many ad’s they can jam between paragraphs. It’s definitely conditioned me to skim instead of read to skip over the superfluous parts that are there simply to extend the length of the article for more ad space.
I will say, I keep coming back here largely in part because of the limited ad’s, but very much because there’s not any auto-play video sucking up my not-free-data connection. I also appreciate your clear and concise article style that doesn’t add a lot of fluff or attempt to please everyone as many publications reach for. I find myself mostly skipping over the PR quotes, and reading your summaries or opinion portions more.
MM
Thankfully, there is still a very reliable way to filter out the “reviews” and “buying guides” that are nothing more than Amazon shopping lists: Look for original photographs. The fake “buying guides”, whether AI or human generated, never have original photos and never demonstrate the tools in use. They just regurgitate stock photos. I suppose it’s only a matter of time before AI fakes its own photographs and generates “deepfakes” of tool use demonstrations but that’s a ways off.
Andrew
It’s already here.
AI is already capable of generative photo and video content, i.e. original.
MM
It’s not good enough to fool anyone yet, unless you’re talking the ultra high end deepfakes which require a lot of human work and I haven’t yet seen used to shill tools. AI generated photos never look right, the people in them are often missing fingers or have extra ones, etc. But sure, at some point that will change.
Steve
I’m actually impressed by how far this technology has come. While I admire Stuart for not using this, I can see many applications for this type of technology, so I don’t think it is going anywhere and will probably cost some their jobs.
College to me seemed to be all about regurgitating information you learned “in your own words.” This always seemed like a waste of time to me to develop 15 pages and make sure it was all formatted and all that. If I could copy some text and have it spit it back out it a different way, I would have been all over that…
Stuart
College experiences can vary greatly, depending on the major.
Much of what I learned has impacted me in meaningful ways. The only class I felt was a waste of my time was when I took economics at the beginning of my freshman year.
I think it was my second year of college when I took a class on early modern Europe. At the start of the first class, the professor explained that the study of history isn’t about the “what, where, or when,” but the “how and why.” Hard science classes were similar.
A lot of people leave college with nothing but a piece of paper. That is indeed a waste of time.
20 years ago, 2 students in the aforementioned history class were caught plagiarizing their term papers. Taking shortcuts to classwork isn’t anything new.
I think that AI is going to change education, where regurgitating information is no longer a testing tactic.
If content generators continue to be a huge problem, teachers and professors will likely move towards more presentation-type assignments, in-class essays, and oral reports.
Collin
>College to me seemed to be all about regurgitating information you learned “in your own words.” This always seemed like a waste of time to me to develop 15 pages and make sure it was all formatted and all that. If I could copy some text and have it spit it back out it a different way, I would have been all over that…
Well, intelligence is (lossless) compression.
Think about that for a moment.
What makes someone intelligent is their ability to losslessly compress all the information presented to them.
The better able you are to losslessly compress the firehose of information that exists in the world, the smarter you are.
For example, a student is given a 1000 page introductory physics textbook. No one can photographically memorize all 1000 pages. The student’s intelligence is directly correlated with their ability to losslessly compress the contents of this 1000 page physics textbook into their mind.
In a way, rewriting information in your own words is a compression exercise. You read a 1000 page textbook. Now can you take the ideas and losslessly compress those ideas into a 5 page essay? Or will you lose some fidelity in doing so?
MM
I agree this is some component of intelligence, but there is also a lot of overlap with wisdom. And there is a lot more to intelligence than simply being able to recall information accurately. One also has to be able to apply it.
When I managed a university lab I worked with many graduate students, postdocs, and professors. Needless to say these people had a great deal of book smarts. But I could tell stories like this all day long:
Example 1: Guy with a BS & MS in Chemistry and a PhD in polymer science decides to use acetone to clean his computer scanner and is shocked when it dissolves the paint and smears it all over the glass.
Example 2: Guy with BS in Chemistry, MS in Materials Science, currently working on his PhD in same is given a classic Ford Mustang by a relative. He decides to start fixing it up. One day he comes in the lab and proudly tells me: I finally got new brake hoses for the Mustang but they were like three feet too long, so I just wrapped the extra hose around the exhaust pipes to keep them off the road.
Steven+B
I’ve worked with AI for decades, professionally. My full time job includes working on pattern recognition software.
Generative AI is garbage. It is just really fancy and expensive autocomplete. I am one of the skeptics. It’s only useful for generating scam content.
At best, it produces plausible-looking content which may or may not be correct…and it consumes a shocking amount of electricity to generate that. If you’re running a business accuracy matters. If you’re paying someone to write something, it either needs to be accurate, like technical writing or instruction manuals or impactful, like sales literature. You can’t risk using generative AI. The whole point is to persuade someone to buy something without breaking the law or ensure a paying customer is using the product correctly and will buy more in the future or not return the item or prevent yourself from getting sued.
Java is much easier to generate than English. The Java code generated looks right until you trace through it and find all the bugs and inefficiencies. It has no clue what it is doing, it is just doing autocomplete based on its training set. It has no clue if the examples are appropriate for your use case.
Finally, I always tell people that if this was so good, Microsoft and others wouldn’t be selling you tools, but selling you finished products. If generative AI actually worked, software is much easier to generate than text. Compilers catch syntax errors. It looks a lot more uniform than text, so there’s much less variation.
If generative AI worked, we’d see Microsoft, IBM, Google, and others selling things like a UI generator for your database…that actually works. Someone would have built a tool that builds your own FPS or RTS video game from prompts. Want a first person shooter? OK, pick the theme, pick the location…maybe even a real-world city…pick parameter, pick your hardware level, etc. You could generate endless content for videogamers and charge for it. For business customers, you could sell expensive cloud services for applications you pay millions for a custom development team to do for you now.
Big Tech hires the best minds and has had decades of low interest, unlimited capital and a huge motivation to solve this problem. They’ve also been working on it for many years. They snatch as many people as possible from academia and pay them salaries they’d be a fool to turn down. If we were anywhere close to it being solved, it would be already and they’d be making a TON of money off the products, not selling you tool kits to build yourself.
Finally, my favorite example is image generation. Every generated image I saw convinced me we are nowhere near replacing human input.
There’s a site called pornpen that generates adult images based on 2 dozen parameters: age, race, hair, general-style, etc. You tell it every detail and it takes 15 minutes (which means a LOT of electricity was consumed to build this), but will generate an adult image for you. When you try it out, it makes a ton of mistakes. Models have extra fingers, if you want more than 2, they get fused like conjoined twins in the most jarring way. They even got the race wrong every time I tried. Even ignoring anatomy errors, the faces are correctly proportioned, but have no texture and look airbrushed. Every AI generated image I’ve seen looks the same…plausible on a bad monitor, but fake on a normal one.
Why do I bring this up? It takes me awhile to read Milwaukee copy or Java code. An image can be judged in a microsecond and every image I’ve seen looked pretty bad unless you had some really strategic lighting. Also, it’s fair to say the commercial internet was built on adult content. There are more freely available pictures of undressed women than anything else on the planet. You can’t imagine a more easy-to-generate image than a woman’s breast with a MASSIVE MASSIVE training set, beyond fathom…and yet it still looks bad. It looks like bad airbrushing.
If they can’t generate a convincing image of a simple form with such a massive training set, what makes anyone think they can solve your business problems?
There’s a reason these technologies are ONLY seen in beta form. If it was any good, you’d be seeing some of the better companies filling the endless demand to have a machine generate things that cost millions of dollars of human time to generate.
Dtree
Can I put this “AI” into the impact gun and how many ugga-duggas is too many?
Doresoom
Milwaukee actually has a few patent applications out there for impact driver profiles controlled via machine learning.
Collin
And by “machine learning” they’re probably referring to a simple linear regression, which is technically an ML technique.
JoeM
Ugh… Nothing is ever going to match the quality standards of Stuart here at ToolGuyd… Okay? It makes me very angry when people start this garbage, just because it’s a slow news day. (Not talking about you, Stuart, I’m talking the people “reporting” about these AI trends, thus triggering you to respond.)
These AI systems that we’re seeing everywhere were never meant to replace Humans, or anything even close to anyone in any field. They were supposed to be written into Social Networking to test specific topics to be generated by an artificial program. Why? They’re segments of a Turring Test. Segments, not the whole. Why can “ChatGPT” do so many different things? It’s because it has to be able to carry on with deceiving human perception as an AI, by being able to catch “Out of Context Banter” in conversation. It’s a mimic of when someone is talking to someone who is an expert in one field, and suddenly ask if they’ve ever tried to master a similar field. Music and Poetry, Graphic Art and Painting, Stock Market Profits and Accounting, the list goes on. Assign this little bastard AI (Programmer’s nickname for a not-quite AI system that is a stepping stone toward the Complete Turring-Class Test AI.) to talk to a huge mass of people, all with different requests of it, and it has to successfully spit out relevant responses.
There are several components to a complete AI, and as such, the bastard AI codes are all tests of each component. Bring them together in one computer system, with the ability to generate a unique visual experience, vocal replication, and knowledge base, enough that their collective entity is able to utterly fool several human beings, in isolated test rooms, tasked with identifying which of the people participating is an AI, and which are Human (With a control group that is all Humans.) and that is where the Turring Test is passed or failed in the current technological standards. Turring limited it to one person, talking on the phone, and the AI is indistinguishable from talking to another Human. With all the forms of communication we have now, video chat, streaming, media sharing… A “Conversation” is significantly more involved. So, without Turring’s limited requirements, we upgraded the standards to include the whole spectrum of who we are as Humans now.
That’s why these AI are supposed to be here. But, yet again, incredibly smart people have let the ignorant chimps play with the Nuke for a while. That’s where we’re entering the major faults, and fear mongering, that has been proliferated by the for-profit Bastard AI systems. Somebody monetized it. Others made it a toy to be played with. At no point did they do the testing they were meant for, so now the Turring Test is failed by default. Now we expect everything to be AI, which leaves absolutely no chance for the Test to be conducted with any control group. It’s infuriating that this goal, set by one of the most brilliant minds of the Computer Age, will never come to pass. All because people just want new toys, and businesses just want to profit off it. Chimps playing with a Nuke. Ignorance always harms the whole, and AI is no different here.
Stuart will always be Stuart. There is no replacing him. Ben V and Benjamen (who we haven’t heard from in a while) were always there to help Stuart, not replace him. Only Stuart can do what he does. Everything else is a Simulation.
Al
“…beyond pro usage environments…”
So, scissors fit for brain surgeons?
Or, scissors for gods? Or at least demi-gods with name-brand sponsors?
Andrew
I think it’s important to keep in mind that this is the worst these AI are ever going to be. They’re just going to continue getting better and better over the years. A good question would be what’s the upper limit for this technology, and I think that’s unclear.
Stuart
I don’t think so.
There are lawsuits over copyright protections, and there might soon be lawsuits about accuracy and misleading answers. At some point AI-generated text and images might need to be labeled as such.
When happens when AI generates false data? Who’s responsible, the person whose prompt generates the data, or the AI service that isn’t careful enough about vetting source materials?
AI can be very powerful.
Right now there’s a mismatch between what it can realistically do and what people mistakenly believe it can do.
10 years ago I wrote a post about how the media was blurring 3D printing reality with science fiction: https://toolguyd.com/3d-printer-capabilities-and-how-the-media-blurs-reality-with-science-fiction-fantasy/
Is there a 3D printer on every desktop yet? Where are the on-the-spot ammunition-printing 3D printers experts were predicting?
Andrew
There are already those sorts of issues with human generate context, all the AI has to do is be better than the humans.
I see your point about 3D printers, and we still don’t have self-driving cars.
However, I’m a bit more cautious about saying it’s not coming, I don’t think that’s clear yet. Computer tech has gone a long way in a very short period of time, and it’s still improving. I’ve got a nicer PC in my pocket than I was able to buy 10 years ago for the desktop.
Who knows where AI will be in 10 years?
Stuart
My point is about the mismatch between what people think AI can do, and what it can actually do.
Generative AIs aren’t synthetic intelligences, they’re simply algorithms designed to draw upon source material to mimic human language.
Will AI produce better content than *some* humans? Yes. And it does this by leveraging existing human-generated content.
MFC
AI is like a useful second brain. Instead of wasting hours trying to think of an acronym you can just ask AI, and while it probably won’t be correct, it will spark your mind to finish the task.
However, I am sooooo sick of obvious fake garbage. They write a six page long essay about scissors that make me want to jump off a proverbial cliff. GIVE ME THE INFORMATION! I don’t want to have to scroll through 6 pages of ads to find that the article isn’t even covering the question I need answered!
But it’s all about money and the longer you stay on their page, the more likely you are to click on something… It would be really nice if Google started offering verified blogs that showed who wrote it, how long the blog has been around, their affiliates and in what country the authors are writing from.
If I’m looking for an answer about nec codes in Texas, then I don’t want to read a blog partially written by someone using AI in Nigeria.
Dave
I wish I could…. I use to write a proposal for capital $$ to do something and was able to get it done with a couple paragraphs.
The latest thing is they set up something like shark tank with a 10 page power point and they want to make a show out of it.
Rx9
As Toolguyd’s resident “John Connor”, it heartens me to hear how bad AI writers are.
I also appreciate your insistence on not using it.
Thanks, Stuart.