I wasn’t familiar with his stuff. Just spent some time on his latest articles.
Great recommendation. Thanks
It’s here for anyone else
Irecently asked Marcus and two other skeptics to predict the impact of generative A.I. on the economy in the coming years. “This is a fifty-billion-dollar market, not a trillion-dollar market,” Ed Zitron, a technology analyst who hosts the “Better Offline” podcast, told me. Marcus agreed: “A fifty-billion-dollar market, maybe a hundred.” The linguistics professor Emily Bender, who co-authored a well-known critique of early language models, told me that “the impacts will depend on how many in the management class fall for the hype from the people selling this tech, and retool their workplaces around it.” She added, “The more this happens, the worse off everyone will be.” Such views have been portrayed as unrealistic—Nate Silver once replied to an Ed Zitron tweet by writing, “old man yells at cloud vibes”—while we readily accepted the grandiose visions of tech C.E.O.s. Maybe that’s starting to change.
If these moderate views of A.I. are right, then in the next few years A.I. tools will make steady but gradual advances. Many people will use A.I. on a regular but limited basis, whether to look up information or to speed up certain annoying tasks, such as summarizing a report or writing the rough draft of an event agenda. Certain fields, like programming and academia, will change dramatically. A minority of professions, such as voice acting and social-media copywriting, might essentially disappear. But A.I. may not massively disrupt the job market, and more hyperbolic ideas like superintelligence may come to seem unserious.
1 Like
That’s a great blurb for Marcus from Kim Stanley Robinson. If you have the inclination, Ministry for the Future is a good, though possibly over-optimistic novel describing our climate future.
Because he beats basically the same drum over and over and is a little long-winded, I take Marcus in small doses. I get it that he can be annoying because he gives himself a lot of credit for being basically right about the direction AI development would take up to now. But hey, as a casual non-expert observer, he was right, afaict. His predictions now? I give them moderate weight. They’re better supported than what the billionaire bullshit artists say.
I will accept there is a craftsmanship factor and techniques involved in the actual cooking that matter but the actual creative aspect of what food to go with what to create a pleasant combination that also balances things like salt fat acid and spice in a good way is the sort of thing I think isn’t that special. Or how to create a diverse menu that appeals to a lot of different people without having too many ingredients in the kitchen. AI can also help support market research, trends, etc. If the restauranter was actually using it to plan all of the things in the way a DM for Dardan or whoever would plan a new chain restaurant I think it would do the job nearly as well.
Anyway it’s a long tangent but in my experience what most people see as creativity is just taking an idea that hasn’t spread to an area yet or taking a new idea and combining it with another new or an old idea in a different way that leverages well understood principles. Basically incremental and evolutionary invention and I think AI can do that as well or better than a really big chunk or the population today. And an expert using AI to do grunt work and to validate their own work can be way more productive and output higher quality than experts who don’t.
I hope we get to your future but I don’t think we are there yet and I’m afraid the cost of getting there may be higher than we want to pay.
2 Likes
This is why the AI discourse is so bad. The “two sides” that dominate the conversation are:
- AI is basically crypto 2.0
- AI is a literal silicon god.
People decide which of those two poles is “more true,” then interpret all news through that prism. Unfortunately, both of those positions are completely absurd! If, in five years, AI is able to economically perform most low-end office work and power robots that can cheaply replace human domestic and industrial labor, but isn’t a godlike superintelligence, is that outcome Team Hype or Team Bubble?
Gary Marcus is a dishonorable exception to the two poles theory, in that he is die-hard committed to both simultaneously. He believes that actually existing AI is crypto 2.0, but that his own AI research program would deliver godlike AI. I don’t know why people want to rehabilitate him just because GPT-5 didn’t score well on benchmarks.
1 Like
The recipe example is a good one to really think through. Unassisted, an AI will not do a good job of producing really delicious and novel dishes. The reason is that the way to create a really good recipe is to start with some known good recipes, add a twist or fuse it with something else, then iteratively refine the product based on experimentation and feedback. An LLM can come up with a reasonable starting point for any dish you care to imagine, but it can’t whip up seven variations, adjust the seasoning, and decide which is the most promising to iterate on. This is why LRMs are so good at math and programming–they can brainstorm promising ideas and then independently try them out to see if they work and use the results to refine their approach.
I find it a lot easier to imagine AI running a research lab for drug discovery than a test kitchen. The loop of brainstorm promising molecule → devise experiment to test efficacy of molecule → run experiment → interpret results is more friendly to AI than the comparable recipe development loop, which critically involves tasting the food to see how good it is.
The really key point here is that intelligence is not some magic universal answer generator that you can point at any problem. Intelligence gives you interesting ideas, but without feedback, refinement, and validation, you can’t reliably think your way very far past what is already known, no matter how intelligent you are.
2 Likes
Ive beaten this drum for a while, but I still think the best use case is going from nothing to mediocre very quickly, in any field
That is astonishingly useful sometimes, and completely (but misleadingly) fucking useless at others
The latter is tech bros thinking they are at the cutting edge of quantum physics or whatever…
On the former. I’m pitching someone for my business on Monday. I’ve spent three hours today intensively learning what a SCADA engineer does, what software and tools they need to know, and a lot of the terms that might come up.
This will be very useful for me as part of my pitch, but my ability to do any SCADA engineering remains at zero.
3 Likes
https://www.reuters.com/investigates/special-report/meta-ai-chatbot-death/
This is beyond messed up.
Elderly man with dementia who had had a stroke ends up chatting to a chat bot. A chat bot designed to lie and flirt
It invites him to a real address, he trips over on the way to meet her and dies. As someone with a dad who regularly makes a mess of Facebook, the transcripts were hard to read.
Zuck apparently made the chat bot team move faster because the versions that didn’t flirt with kids were too boring.
These people are just straight up super villains at this point.
4 Likes
trust me you dont wanna read the ghoulish comments on tech forums about it
Paraphrase them for me?
“the chatbot had nothing to do with him falling, he could’ve fallen regardless”
2 Likes
Anyone remember T-shirt art from the 90’s or 80’s that showed a 3x3 grid of pictures showing one character morphing into another character? I’m looking for a friend who wants to try and get an AI to do one for her but neither of us has had much luck even finding examples of it let alone getting an AI to generate one. Any help on prompts would be appreciated.
You mean like animorphs?
Possibly? I didn’t really remember them so I’m going of my friends description. Chatgpt thought they were “Lenticular Morphs” but the closest thing that got me at artist.io was this:
Which isn’t 9 and not really morphing from Charlie Brown into Himmler (don’t ask).
Zero reason chatbots should flirt and all kinds of reasons why they should not.
1 Like
One reason they should flirt
An adult signs up to a “flirt with me” chat bot. Explicitly.
I would argue nobody needs that even for practice.
