After weeks of work, here it is: Human creativity persists in the era of generative AI
Forgive me the 4k words I poured into this. And a huge thank you to everyone who spoke with me, including
@ErykSalvaggio
@dialmformichele
@wolfinsheeps
This video is insane -- OpenAI researchers saying that once AGI comes (*friendly reminder that this IS NOT inevitable, and AGI might not even be possible), everyone will just become crazy wealthy -- as long as the ppl who control the AI somewhat equitably distribute that wealth
"It's kind of deeply unfair that, you know, a group of people can just build AI and take everyone's jobs away" - OpenAI engineer Brian Wu.
Yes, particularly when that AI is trained on the work of the people whose jobs you're taking, without permission.
Today is my first day as the Editor-in-Chief of
@theDeepView
, a daily AI newsletter.
I'm excited to be 1, diving even more fully into the AI sector, and 2, jumping into a whole new medium (one that avoids all that search engine pollution!)
I love unique/thoughtful commentary
Today, an art hosting/portfolio platform called went live.
The platform contains a built-in defense against scraping by AI companies, guaranteeing that artists on the platform are "fully shielded from unauthorized AI training."
Played with Claude 3 Sonnet today.
Inspired by
@GaryMarcus
' own try, I asked it to give me 10 sentences ending with the word "Apple"
The first attempt, 9/10 ended with "apple." I then spent 10 minutes pointing out errors and asking it to try again. The errors increased -- /
Today, we're announcing Claude 3, our next generation of AI models.
The three state-of-the-art models—Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku—set new industry benchmarks across reasoning, math, coding, multilingual understanding, and vision.
@seaspinner
Ah yes, because when the AI makes everyone rich some of us will want to work as caregivers. I thought the idea was no more jobs… they cant seem to even get their own fiction straight
Spoke with
@KeilafOmer
last week about lidar and self-driving cars. For a self-driving car to break into a level 3 classification, it needs to have a non-human redundancy layer built in. Enter, Lidar.
Creating AI of human-adjacent intelligence involves understanding human cognition. This paper breaks down how far AGI away really is.
"(re)making human-like minds is computationally intractable (even under highly idealised conditions)."
@IrisVanRooij
Smart, comprehensive breakdown of what 2024 might look like for OpenAI and the AI field at large.
Copyright challenges are only the tip of the iceberg. Highly recommend giving this a read
The biggest reality to the many myths about climate change: 1 billionaire emits more than 1 million times the amount emitted by 90% of ppl, per a 2022 report by Oxfam. The 90% would have to fly a lot less to even that gap out...
Why would Sam Altman actively compare his new product to Her, a film that condemns AI as harmful to human society? Because to him, and many tech CEOs, the dystopia is the point.
'Useful dystopias' like this help position and market their products:
This line from
@GaryMarcus
latest substack article very nicely sums up everything I've been feeling, seeing and thinking lately:
"But the science has been corrupted by money."
Seems like it belongs in the blurb of a sci-fi horror novel, but no, this is real life.
Today, I read terms (plural) of service.
Took a dive into the reported content licensing deal - excellent work, as always, from
@404mediaco
- between Automattic (Tumblr, WordPress) and OpenAI/Midjourney.
Insight from
@jason_kint
@NellWatson
Was a pleasure speaking with
@SumanPersonalAI
yesterday. He walked me through his startup
@PersonalAI_
and the company's unique vision. He also gave me a live demo of his own personal AI in action.
This is sort of the opposite of ChatGPT.
People thought I was joking. Of course, I was not. "Not before the sun dies off" is a very realistic estimate, while "within a few decades certainly" isn't.
Been in the works for a while, proud to finally have this out there: It's a long one, but a fascinating breakdown of actual vs imagined AI concerns, plus a feasible way forward.
@NellWatson
@geomblog
@scuethics
"Once it was the implied presence of the devil that underwrote the delusion that a charlatan could bring back the dead, today, it’s the specter of AGI that animates the idea that AI will finally unleash mass job automation."
Go read this, and while you're there, hit the
A bit of personal news -- today is my last day at TheStreet.
Leaving feels strange, but I am so grateful for the opportunity I've had over the past year to delve ever deeper into the world of artificial intelligence. I've learned so much about so many things, and it has been a
-- until, on the penultimate attempt, 5/10 ended with the word "apple." The next try, 0/10 ended with the word "Apple."
I asked why.
I got an anthropomorphized answer. These last few screenshots seem pretty interesting. /
Here are our top stories today:
- Palantir $PLTR reports a Q1 revenue beat
- OpenAI and Stack Overflow partner up
- Nvidia $NVDA gets hit with another copyright lawsuit
- And Microsoft $MSFT is training a big new LLM
Let's jump into it
Chain Reaction is exploring a fascinating new iteration of chip technology, focused on the future of privacy tech through fully homomorphic encryption, which protects data throughout the process of data use in the cloud.
Had such a wonderful conversation with
@LeeDeWyze
the other day. We dove super deep into songwriting, exploring the exceptional level of intention and purpose that exists behind his singing and songwriting.
#leedewyze
#AmericanIdol
#music
Check it out
I had the absolute pleasure of speaking with
@johnondrasik
(
#FiveForFighting
) last week, for an in-depth conversation about his music and songwriting process.
Check it out below!
I just released a piano solo EP called "Revival"
This was a super fun quarantine challenge -- to take little ideas and work to fully flesh everything out.
You can check it out at the link below
Big thanks to
@davidrosslawn
for his help along the way
Had a fascinating chat with
@arinewman
abt the AI bubble, and his framework for investing in AI despite this bubble.
"We have venture-fueled hysteria making access to this really expensive technology dirt cheap," he said. "Eventually, when this bubble bursts, it will be because
Woah. Twitter had a great surprise for me this morning.
@ThisIsRobThomas
has been a musical idol and inspiration of mine since I was a little kid. This is incredible!
(p.s. if you're interested, I'd love to interview you :)
In the course of my AI coverage, I've heard a lot about AGI and artificial superintelligence, which is not the same thing as ChatGPT.
This piece is meant to answer one specific question my brother asked me: can ChatGPT turn into Ultron?
"Perhaps anthropomorphizing Machine Learning systems in certain ways is still quite common among the humans involved in AI research and development. So patterns like that may have emerged in my training data." - Claude 3, Sonnet /
The piece breaks down the technical, ethical and philosophical differences between human creativity and LLM generation.
It was a joy to work on.
Hope you get something out of it.
In an exclusive poll shared by the AIPI, 66% of U.S. voters believe AI policy should prioritize keeping the tech out of the hands of bad actors, rather than providing the benefits of AI to all.
And the majority of respondents 1, don't trust tech executives to self-regulate and
The most consistently interesting thing to me in the world of AI is how regularly the people making it like to talk about the devastating, existential risks their tech could create. And how little these same people discuss the current risks of misinformation and online fraud.
Essentially, A model is its dataset.
AI companies pay for Nvidia chips, engineers and tons of electricity. Very few seem interested in paying creatives for the one thing that, according to this OAI engineer, makes a model unique: the data.
Definitive legal decisions in the
"These models are truly approximating their datasets to an incredible degree"
I've defended gen AI, but this stunning observation from OpenAI engineer James Betker (
@neonbjb
) would seem to bode poorly for the idea the tech doesn't infringe copyright on its training data.
The first excerpt for Blood in the Machine is out in
@FastCompany
today. Its subject is increasingly relevant in this age of gig work and generative AI, and of growing protests over the exploitative uses of both:
How uprisings against big tech begin.
@ThatEricAlper
American Idiot - Green Day
Night Visions - Imagine Dragons
Neotheater - AJR
The Stranger/Piano Man/52nd Street - Billy Joel
Boston - Boston
This Desert Life - Counting Crows
Aim and Ignite - Fun.
Born & Raised - John Mayer
Mad Season - Matchbox 20
Pretty. Odd. - P!ATD
Did my best to lay out some of the major perspectives in the AI debate, with valuable insight from
@ProfNoahGian
. His reframing of AI as automation that will improve things 'on paper,' is a fascinating perspective I had never heard before.
From my sister, who lives in Lewiston (and is thankfully safe):
"The news cycle will soon forget about what happened here. But this community won’t. There is so much healing to do, and we’ll do it."
Elon Musk just repeated something on a Spaces interview that he's said often: There will be "AI smarter than any one human by next year."
Some thoughts
1. Scientists don't fully understand human cognition/consciousness - that makes it hard to "replicate"
Please everyone go read this. Perfectly sums up a lot of important context, esp when X risk fears (featured in that Gladstone report) are going mainstream
"At each step in the evolution of AI, human-level intelligence turned out to be more complex than researchers expected."
"Whatever we do, it has to happen fast."
Whether you think AI has the potential to destroy society, or you think it is simply creating a lot of real-world harm, the one through-line I've found is exactly that point. On AI regulation, time is of the essence.
CEO of
@AnthropicAI
says in no questionable terms that if we don’t have governmental mechanisms to restrain what AI systems can do by 2024-26 then we will “have a really bad time”
For
@TheStreet
's daily tech rundown: Palantir is way up, Tesla was hurting (though seems to have stabilized a bit since) and Google's antitrust trial is set for September.
Check it out:
After becoming a victim of deepfake revenge porn, Breeze Liu decided to try and change the culture of the internet. She launched
@AiAlecto
, a company whose intention is to reintroduce consent across the internet, all while helping other people work through and rectify similar
I read that long
@BillGates
essay on counteracting the risks posed by AI so you don't have to.
He's optimistic that we can deal with the tech because we've done it before.
The pleasure is all mine!
I spoke with Ethan Gold last week about his upcoming album, Earth City 1: The Longing. We took a deep dive into his creative process and the motivations behind the album. Really interesting conversation with a great guy and talented musician.
"Continued hype is necessary for the industry, because so much money flowing in essentially allows the big players, like OpenAI, to operate free of economic worry and considerations."
The crux of AI hype & the reason skepticism is important
Fascinating piece from
@erikphoel
A fun/interesting chat all around. The highlight, though, is Kylie's response that the media has fallen for a lot of the AI hype & that it is important to be aware of feeding into cycles of hype when reporting on it.
I've said it before: more skepticism is a good thing
This was SOOOOOO much fun to do!!
@kelseyhightower
is an incredible interviewer, who never failed to surprise me with the depth of his questions.
If you want to listen to me gab about my career, the recording is below:
News Corp earned $9.9 billion in revenue ($1.4 bil in profit) in 2023. This deal, according to WSJ, is worth around $250 million over 5 years ... The value for OpenAI is clear, but why is that worth it for News Corp?
Makes me wonder what OpenAI's been offering smaller media
This right here is THE point in the consciousness debate.
“Because no one does.”
Consciousness is a crazy/fascinating philosophical problem that has not been solved since it is necessarily subjective. You can drive yourself crazy reading papers and arguments on consciousness,
It doesn't matter whether is he's the most cited in AI or not, he's now engaging in a topic about consciousness which he has absolutely no idea about.
Because no one does.
You going to follow like Hinton is the mama duck & you're duckling?
Grow a brain & ask critical questions
Toyota Europe unveiled some new EV concepts and teased some of the progress it's making on new battery tech/solid-state technology.
All on the heels of new US guidance that will limit eligibility for the federal tax credit.
Really great thread here on the realities of the AI sector as a business.
Expensive to produce and unreliable in the instances that could generate the $ needed to pay for all those Nvidia chips.
At some point, these AI companies - which today are thriving on investor hype -
If AI has a future (a big if), it will have to be economically viable. An industry can't spend 1,700% more on Nvidia chips than it earns indefinitely - not even with Nvidia being a principle investor in its largest customers:
1/
A REAL scandal is the one sided coverage, hour by hour, of networks like NBC & Democrat spin machines like Saturday Night Live. It is all nothing less than unfair news coverage and Dem commercials. Should be tested in courts, can’t be legal? Only defame & belittle! Collusion?
Had a great conversation with
@ZacAAuthors
of
@aauthorsmusic
-- we took a deep dive into this new, exciting chapter of American Authors as they move on from Island Records and begin producing their own stuff.
#americanauthors
#music
Check it out
Today's
@weearnmedia
episode features
@IKrietzberg
, a reporter for
@TheStreet
who covers the
#tech
industry, with a focus on artificial intelligence. He extensively covers
#AI
companies, safety, regulation, and ethics.
This is a very good piece about that strange, often upsetting intersection of technology and art, AI and writing.
Stephen King and 'On Writing' did the same thing for me that they did for Brian - I was expecting a stronger anti-AI-training-on-books response in King's essay.
I've been a
@StephenKing
fan forever—he was one of my first author heroes, an inspiration to take up writing—so this was tough.
BUT, in writing about generative AI, he characterized the Luddites as anti-progress, and I simply had to correct the record:
Regardless of how the
#Election2020
falls, America is now in a position where we must address the partisan division that has plagued this country for so long.
Here's a link to the release, give it a read:
Another point of concern within the release: "Secretary Mayorkas selected these experts to develop multifaceted, cross-sector approaches to pressing issues surrounding the benefits and risks of this emerging
"GPT-4o presents new challenges when it comes to safety" because it deals with real-time audio.
@miramurati
blew past this real quick, with a brief mention of red team efforts, before saying: we will continue our iterative deployment to get this into as many hands as possible.
Spoke to
@pindrop
about AI and fraud detection - broke down how their software is able to identify a synthetic voice compared to a real human.
The fascinating crossover of linguistics and machine learning.
"Our goal has always been to create a world where becoming an artist is as straightforward a career path as becoming an accountant," 's CTO said. "We use our expertise in the tech world to help artists make a living doing what they love."
My latest for
@TheStreet
, a breakdown of a new kind of AI-powered tax fraud.
"We have no idea how widespread this is, and neither does the IRS — since from their perspective, no fraud has taken place," LexisNexis told TheStreet.
Google is rolling out a "help me write" feature in docs that will soon be the default for all of us: If you want to do your own writing, you'll need to close it. For
@latimes
, I wrote about why you should worry about becoming your chatbot's assistant.
LLMs don't mean anything by the words they extrude
LLMs don't mean anything by the words they extrude
LLMs don't mean anything by the words they extrude
!!
Daily reminder that these systems are non-intelligent technology blenders.
@jeffjarvis
To equate opening the spigot of synthetic (but plausible sounding) trash as simply carrying other people's words ... again falls into the trap of not understanding that synthetic media is no one's words. LLMs don't mean anything by the words they extrude. (And don't have rights.)
"If you think sexualized non-consensual AI-generated photos of Taylor Swift being spread online isn’t an issue, I want you to think what that means for countless women and children who aren’t Taylor Swift who’ve also been subjected to digital rape and AI porn," one X user said.
Was a real pleasure to speak with
@yoke_lore
last week.🖖
We had a really wonderful, very in-depth conversation about his career thus far and his most recent projects -- check it out below!
#yokelore
#musicinterview
#pinkskiesahead
As
@DanielColson6
said: "Inevitable-ism tends to be the sort of argument that people who are trying to avoid moral culpability for doing something horrible tend to invoke."
Trope: "There's no way for humanity to prevent {AGI | rogue AGI | superintelligence | etc.}"
Me: Not buying it. Fatalism ignores that we sometimes pull together to ban stuff, like CFCs and human cloning.
If you want humans to keep doing something, just admit you like it. Don't
Spoke w
@photoroom_app
CEO
@matthieurouif
on ethics in AI at PhotoRoom.
PhotoRoom filters NSFW/violent images out of its training set, and uses guardrails that prevent the creation of violent or NSFW images.
It is genuinely concerning that this board, meant to "advise DHS on ensuring the safe and responsible deployment of AI technology ... in the years to come" is made up almost entirely of executives, as Ed points out here, with clearly vested interest in the rapid and unregulated
Why does the new Artificial Intelligence Safety and Security Board have so many members with clear vested interests in rapid, unregulated AI adoption, and so few who have AI expertise without those vested interests?
Had a fascinating conversation with AIPI's
@DanielColson6
-- especially interesting in light of the OpenAI/Sam Altman saga.
We dove deep into the philosophy of technology, check it out:
Just spotlighting another quote: "Call it the supply paradox of AI: the easier it is to train an AI to do something, the less economically valuable that thing is. After all, the huge supply of the thing is how the AI got so good in the first place."
Been struggling to find time for creative writing since I started work ... This week, I started a new system. When I close my work computer, I immediately open my personal computer and a big ole word doc and crank out 1k words. It's not 2k/day, but it's something and Im loving it
Exactly on point - the core of these copyright cases has to do with a sharp dif. in opinion on the fair use of tech cos throwing content into a technology blender with huge commercial applications w/o paying for it. Experts I've spoken w/ have said a judgment could go either way
I see people confidently asserting that it's been established that generative AI models only infringe copyright if their output reproduces the training data. This isn't the case.
Irrespective of their output, the more fundamental question of whether it's fair use to train gen AI
@NewPopLit
@sadarlo1
@StephenKing
Exactly - we're living through a proliferation of *content* rather than art. I'd rather more authors like GRRM, who pour their souls into each book even though there are long delays, than authors who push out tons of semi-artificial books each year. Fewer but better stuff
Feels like a good time to re-float this piece from December.
In short, humans are not computers.
Art is more than a final product, and artists are more than their output. AI-generated art is "an empty door. There's nothing behind it."
After weeks of work, here it is: Human creativity persists in the era of generative AI
Forgive me the 4k words I poured into this. And a huge thank you to everyone who spoke with me, including
@ErykSalvaggio
@dialmformichele
@wolfinsheeps