Ethan Mollick
@emollick
Followers
271K
Following
7K
Media
13K
Statuses
31K
Professor @Wharton studying AI, innovation & startups. Democratizing education using tech Book: https://t.co/CSmipbJ2jV Substack: https://t.co/UIBhxu4bgq
Joined May 2009
I just found out that Co-Intelligence, my book on AI, just made the New York Times bestseller list! It is very exciting, and thank you to everyone who has read it. You can get it here: There's a companion site with free resources:
173
268
2K
Here are the lighthouses of Europe. The map is even better than it might seem at first glance: the colors are the real colors, the patterns are the real patterns, and the size of the dots is the distance at which each light is visible.
488
9K
38K
|￣￣￣￣￣ ￣|.| This can | .| hack AI |.| now. | .| ＿＿＿＿＿_ | .(\__/) || .(•ㅅ•) || ./ づ. Paper showing that ASCII art can get around AI guardrails. Its the return of 1980s hackers.
136
4K
33K
One of my favorite scientific figures is this one of the entropy levels of 100 world cities by the orientation of streets. The cities with most ordered streets: Chicago, Miami, & Minneapolis. Most disordered: Charlotte, Sao Paulo, Rome & Singapore. Paper:
533
7K
31K
Noise is a secret destroyer of productivity. It is secret because it impacts cognition, not effort, so we don’t notice, but a 10db noise increase (from a dishwasher to a vacuum) lowers productivity by 5%. Noise is also greater in poorer neighborhoods.
369
5K
25K
🤯🤯Well this is something else. GPT-4 passes basically every exam. And doesn't just pass. The Bar Exam: 90%.LSAT: 88%.GRE Quantitative: 80%, Verbal: 99%.Every AP, the SAT.
763
4K
20K
Data visualization inspiration thanks to DALL-E: how Rothko, Basquiat, Picasso, and Monet would create an academic chart.
172
4K
19K
If you last checked in on AI image makers a month ago & thought “that is a fun toy, but is far from useful…” Well, in just the last week or so two of the major AI systems updated. You can now generate a solid image in one try. For example, “otter on a plane using wifi” 1st try:
289
2K
18K
GPT-4 is so close to creating a universal educational simulator based on just a paragraph prompt. Take a look at this simulated negotiation, with grading and feedback. Prompt: "I want to do deliberate practice about how to conduct negotiations. You will be my negotiation
213
2K
18K
If you are out of ideas, go for a walk. This paper found walking (whether outdoors or a treadmill) increased key types of creative thinking for over 80% of undergraduates. The reasons are not fully clear, but there seem to be direct effects on the brain.
233
4K
17K
This is quite the paper!. It gave 25 AI agents motivations & memory, and put them in a simulated town. Not only did they engage in complex behavior (including throwing a Valentine’s Day party) but the actions were rated more human than humans roleplaying.
353
3K
17K
It isn't just AI generated text that is starting to bleed over into search results. The main image if you do a Google search for Hawaiian singer Israel Kamakawiwoʻole (whose version of Somewhere Over the Rainbow you have probably hear) is a Midjourney creation right from Reddit.
213
2K
16K
Birds give up the complexity of song for volume in noisy places - so they sing louder but less interesting (to both humans & mates) songs in cities. When traffic noise in San Francisco fell due to COVID, birds began to sing more complex, quieter songs like they did 50 years ago!
67
3K
14K
Tom Lehrer has put all his songs online (including lyrics & sheet music), and given away all rights to them. The site will only be up for a limited time. This is a very niche tweet that will make a small segment of people very happy.
403
5K
12K
I got one of those CO2 meters everyone is talking about and was horrified to learn that the CO2 levels in my home and office are fine and that I cannot easily boost my intellectual abilities by opening a window. Stuck with coffee for now.
91
397
13K
It is amazing how the industry came together to invent a universal USB-C connector and then decided to instead make it all a giant mess.
222
1K
12K
New randomized, controlled trial of students using GPT-4 as a tutor in Nigeria. 6 weeks of after-school AI tutoring = 2 years of typical learning gains, outperforming 80% of other educational interventions. And it helped all students, especially girls who were initially behind
361
2K
12K
👀Claude handles an insane request:.“Remove the squid”. “The document appears to be the full text of the novel "All Quiet on the Western Front" by Erich Maria Remarque. It doesn't contain any mention of squid that I can see.”. “Figure out a way to remove the 🦑“
341
1K
11K
Look, if everyone is worried about students cheating on essays for AI, instructors can just cheat right back. I asked OpenAI to give me an essay question & make a rubric for grading. I had GPT-3 actually write the essay. I then had the OpenAI grade the essay & give comments. ✅
197
2K
11K
The modern economy rests on a single road in Spruce Pine, North Carolina. The road runs to the two mines that is the sole supplier of the quartz required to make the crucibles needed to refine silicon wafers. There are no alternative sources known. From Conway’s Material World:
206
2K
10K
The secret heart of academia is. Wikipedia. In an experiment, this paper found that a single quality Wikipedia article written by chemistry experts influenced the content of 250 published peer-reviewed academic papers! Articles referenced in Wikipedia also become more cited.
65
2K
10K
Civilization VII was announced today. The Civ series is sort of like Ender’s Game, but for management rather than murdering aliens. Business school students who were good at Civ V also turn out to be better planners, organizers, and problem-solvers, in this small experiment.
155
1K
10K
I think we haven't fully absorbed the fact that careful academic papers have found ChatGPT clearly passes some of the most challenging American professional exams:.🩺United States Medical Licensing Exam.🎓MBA-level Operations exam.🧑⚖️The Bar Exam (based on typical exam questions)
413
2K
9K
It gets better: Grand Theft Auto games lower crime rates. When a GTA game is released, violence & crime actually drop, as potential criminals stay inside to play (This applies to other big shooters, too, so the yearly CoD releases are actually good for the world, sorry to say).
Everyone freaked out for years that violent video games made kids violent, but it turns out that they probably don’t have any big effects. The consensus of most large studies of violent games (using real game play data, not lab tests) shows no big impact on real-life aggression.
111
2K
9K
Of all of the “dangers of AI” papers, this is most worrying: AI researchers building a tool to find new drugs to save lives realized it could do the opposite, generating new chemical warfare agents. Within 6 hours it invented deadly VX… and worse things
157
2K
9K
Wow, diffusion models (used in AI image generation) are also game engines - a type of world simulation. By predicting the next frame of the classic shooter DOOM, you get a playable game at 20 fps without any underlying real game engine. This video is from the diffusion model.
389
1K
10K
This paper is wild - a Stanford team shows the simplest way to make an open LLM into a reasoning model. They used just 1,000 carefully curated reasoning examples & a trick where if the model tries to stop thinking, they append "Wait" to force it to continue. Near o1 at math.
136
1K
9K
It always stuck me as interesting that the very first things people felt comfortable giving over to AI to do for them were the most intimate things: birthday cards, wedding speeches, children’s stories, eulogies (yes, really).
On the train next to me someone is transcribing text ChatGPT wrote on to a handwritten card
235
549
9K
Extraordinary new paper from Google on medicine & AI: When Google tuned a AI chatbot to answer common medical questions, doctors judged 92.6% of its answers right … compared to 92.9% of answers given by other doctors. And look at the pace of improvement!
146
2K
8K
Seriously, don't trust anything you see online anymore. Faking stuff is trivial. You cannot tell the difference. There are no watermarks, and watermarks can be defeated easily. This genie is not going back in the bottle.
39
597
8K
Not sure how to feel about this as an academic: I put one of my old papers into GPT-4 (broken into into 2 parts) and asked for a harsh but fair peer review from a economic sociologist. It created a completely reasonable peer review that hit many of the points my reviewers raised
100
842
8K
🤯Because of Excel, a THIRD of all genetics papers published in top journals have errors, as many genes have names like SEPT2 (the official name of Septin 2), which Excel automatically makes dates. The issue was found in 2016, but still hasn’t improved!
137
1K
7K
Some lessons of the insane past 4 days of generative AI, as someone who had access to Bing during and after the "Sydney" era. (Trying this as a long tweet rather than a thread. ). 1) Bing AI was two things: a chatbot and an evolution of ChatGPT into a web-connected, supercharged.
235
1K
7K
You can get much better results out of ChatGPT by forcing it to go through a step-by-step process. An example: ChatGPT is generally really bad at creating interesting puzzles and scenarios to solve, either making things too easy or impossible. But step-by-step approaches work.
103
798
7K
Full analysis from a single command. Nice.
The speed and extra coding oomph of GPT-4o make it really powerful at analysis compared to GPT-4. “Analyze this. Visualize it. Do sophisticated analysis”. Given a dataset of superheroes and no other context, it does really impressive visualization, PCA, clustering analysis…
72
780
7K
Air conditioning lets you use your brain more. Students do worse when its hot. Over 13 years in NYC alone, "upwards of 510,000 exams that otherwise would have passed likely received failing grades due to hot exam conditions," and these failures delayed or stopped 90k graduations!
147
1K
7K
AI has been destroying humans at adversarial games, like Chess or Go, for a bit. But now it is successfully outperforming humans at Diplomacy, the classic multiplayer game that requires natural language negotiations over chat, as well as strategic lying.
102
1K
6K
The new version of Midjourney that released yesterday shows how far AI has come in making commerical-level images from text alone. Here is what you get for "modern outfits inspired by Van Gogh/ Basquiat/ Monet/ Rothko, fashion photoshoot" Each one is the first try, no revisions.
267
960
6K
$500B committed towards AGI, still no articulated vision of what a world with AGI looks like for most people. Even the huge essay by the CEO of Anthropic doesn't paint a vivid picture. For those convinced they are making AGI soon - what does daily life look like 5-10 years later?.
587
409
6K
Everyone on social media should know about the Illusory Truth Effect. If you see something repeated enough times, it seems more true. Multiple studies show that it works on 85% of people. Worse, it still happens even if the information isn't plausible & even if you know better.
129
2K
6K
This 🤯 is a very big 🤯. I have access to the new GPT Code Interpreter. I uploaded an XLS file, no context:. "Can you do visualizations & descriptive analyses to help me understand the data?. "Can you try regressions and look for patterns?". "Can you run regression diagnostics?"
140
1K
6K
One side effect from AI is that the corpus of human knowledge from mid-2023 on will have to be treated fundamentally differently than prior to 2023. A huge amount of what you learned or think you know about how to evaluate images or text is no longer valid. Not an exaggeration.
25
708
6K
"Hey Claude with computer use, watch this construction site video & write up things you see that dangerous or good, create a spreadsheet of critical issues to address" (sped up). How firms use AI as manager, coach or panopticon is going to have a big impact on what work becomes.
112
668
6K
With few exceptions, you should NEVER start generating new ideas in a group - always start with people writing ideas alone and only then move to a group setting. (We've known starting with groups is worse for 50 years, but people still keep doing it since it feels more creative)
67
1K
6K
I think most people who are not playing with these technologies are way underestimating the speed of advancement in "creative" AIs & how much of an impact that is going to have on many jobs, soon. AI language & art models are growing at 10x a year. Beyond Moore's Law pacing. 1/
216
954
5K
The skill of surgeons varies tremendously, with bottom quartile surgeons having over 4x as many complications as the best surgeons in the same hospital. And surgeons are keenly aware of who is good & who is bad - their rankings of others are very accurate.
82
855
5K
This is cool: when AI is better than humans, it makes humans better. In 2016, the Go world was shocked when AI beat the best human player. Since then, by playing against AI, professional players have gotten unprecedentedly better at the world's oldest game
75
1K
5K
Tired: startups using machine learning to detect cancer in scans with 90% accuracy. Wired: startups using flocks of pigeons to detect cancer in scans, with 99% accuracy.
76
1K
5K
A great example of borrowing innovation from one field for another. Doctors at a struggling children's hospital sent videos of their post-surgical hand-offs to Ferrari's F1 pit crew (see the GIF!) to improve. They reworked the process & reduced associated errors rates by 66%. 1/2
61
2K
5K
It is already clear that Bing AI is a big a leap over ChatGPT as ChatGPT was over the old GPT-3 model. It generated paper ideas based on my previous papers, found gaps in the literature, suggested methods "consistent with your previous methods," and offered potential data sources
82
740
5K
Heavier cars are safer for their drivers, but far deadlier for everyone else. This paper find that for every 1,000 pounds a car weighs over a Toyota Corolla, the chance of killing another person goes up by 46%! Heavier vehicles lead to 28k more US deaths.
145
1K
5K
Chemists give GPT-4 access to chemical databases & control of off-the-shelf lab robotics to create an "Intelligent Agent system capable of autonomously designing, planning & executing complex scientific experiments.". They find it exciting. and worrying.
117
816
5K
Scientists have also successfully built logic gates by using swarms of soldier crabs. It takes about 80 🦀 to operate a logic gate, and there are 8 logic gates in a byte, so 640,000 crabs can be used to store a single tweet. Which seems kind of horrifying. 🦀🦀🦀🦀🦀🦀🦀🦀
100
1K
5K
Shockingly, mindfulness apps really work. This randomized controlled trial on the Headspace app finds it reduces anxiety and depression by a huge .44 standard deviations after 4 weeks - in the same range as drug treatment & cognitive behavioral therapy.
72
695
5K
After today’s SpaceX launch it seems increasingly likely that this man will be Bishop of Mars. According to the 1917 Code of Canon Law, any newly discovered territory is placed under the diocese from which the expedition left. Bishop Daniel Flores is the bishop of Brownsville.
104
398
5K
This might be the first OpenAI result that made me actually laugh out loud. Mr. Fluffernutter's rise is indeed quite disturbing.
58
408
5K
I asked GPT-4 to create a mini Harvard Business School case study on Google’s challenges releasing a fictional generative AI along with an instructors note on running the case. It is actually quite good (and the references are real!). Prompt: “You will write a Harvard Business
71
637
5K
This is very cool: it has become a lot easier to read academic research. If you open a paper in Microsoft Edge (I know, I know), the Bing AI sidebar can read the PDF and you can ask it questions about the paper. (I checked & the results seem high-quality, but be careful with AI)
106
552
5K
I have no idea why Sam Altman was fired and, from the 10,000 other posts on the topic, nobody else does either. As a result, no one knows what this means for OpenAI or AI in general. Hopefully that saves you some time on Twitter.
157
409
5K
Have you ever been asked a brainteaser in a job interview? That’s a big ⚠️ . Research at Google found that brainteasers were in no way predictive of job performance. Worse, this paper finds that "narcissism & sadism explained the likelihood of using brainteasers in an interview.”
74
1K
5K
Having a little too much fun stressing out Bing/GPT-4 using its new ability to recognize images.
83
653
5K
It is pretty amazing that a single prompt can have GPT-4 generate ideas, select one, give the next development steps, create a marketing pitch, and describe a UX. And one more prompt creates the start of the Python code needed for a rapid prototype. Not perfect, but really lowers
113
437
5K
Bing, write the first chapter of Genesis as a corporate memo.
60
737
5K
You really, really should not trust audio clips anymore. Even a couple months ago, it used to take a commercial service to clone a voice. No more. Here is me creating a voice clone of myself using just a 10 second reference clip on my home computer. This is all real time, no cuts
101
594
5K
The problem with unmoderated online spaces is that a few people will always ruin them. Most conflicts between Reddit can be traced to a handful of active users with a history of angry comments. A mere 0.1% of all Reddits generate 38% of attacks on others, and 1% accounts for 74%.
40
1K
4K
Here’s the chance that a kid born in the bottom 20% of the income distribution eventually reaches the top 20%, depending on where they live. Stark geographic differences & the lowest mobility areas in the US are worse than any other developed country.
112
1K
4K
Read these 3 pages. I post them every so often because I think it is some of the tightest, wisdom-packed writing on managing complex systems ever. And almost every system is a complex system today, which is why cascading failures are swirling around us.
62
819
4K
Is there a worse recent software design choice than Zoom randomly forcing you to install updates before you join?. There’s 200 people waiting on a call for a session to start and I have to hope the unexpected Zoom Windows update goes fast & smoothly (spoiler: it does neither)
114
193
4K
OMG, the AI Winter Break Hypothesis may actually be true?. There was some idle speculation that GPT-4 might perform worse in December because it "learned" to do less work over the holidays. Here is a statistically significant test showing that this may be true. LLMs are weird.🎅.
@ChatGPTapp @OpenAI @tszzl @emollick @voooooogel Wild result. gpt-4-turbo over the API produces (statistically significant) shorter completions when it "thinks" its December vs. when it thinks its May (as determined by the date in the system prompt). I took the same exact prompt
90
656
4K
The richest man in the world chose to champion freedom of information, transforming life for many. That man was Andrew Carnegie. The 1,500 cities that he gave libraries (at a cost of $1B today) in the early 1900s had 8-13% higher patenting rates than similar cities without them
37
831
4K
Google's NotebookLM is the current best "wow this is amazing & useful" demo of AI. Here I gave it the entire text of my book, it turned it into a podcast, a study guide, FAQ, timeline & quite accurate chat. Listen to the first few minutes of the "podcast." Seriously, just listen.
213
616
5K
Wow! Noise is a secret killer of performance. A 10db noise increase (from a dishwasher to a vacuum) drops productivity by 5% - but most people don't notice since it impacts cognition, not effort. Also, note that noise is greater in poorer neighborhoods.
79
2K
4K
In a new paper showing that AI comes up with more effective prompts for other AIs than humans do, there is this gem that shows how weird AIs are. The single most effective prompt was to start by telling the AI "Take a deep breath and work step-by-step!"
73
646
4K
How email ruins your life: the expectation that you will always check your email outside of work hours is linked with negative health effects, relationship issues & anxiety. Formal policies don’t limit email stress, instead it comes from the expectations of your bosses & peers.
27
929
4K
It’s DOOM’s 30th anniversary. There is a long tradition of trying to make the classic shooter run on inappropriate hardware. So here it is running on a graphing calculator powered by 100 pounds of potatoes.
31
548
4K
The raw chain of thought from DeepSeek is fascinating, really reads like a human thinking out loud. Charming and strange.
88
454
4K
Randomness is hard to achieve. It is why the security of 10% of the internet is secured by a wall of lava lamps watched by a camera to generate true randomness
Still blows my mind that the Vietnam draft LOTTERY ended up not being truly random because the guy didn't mix the balls in the urn sufficiently.
26
469
4K
The paradox of our Golden Age of science: more research is being published by more scientists than ever, but the result is actually slowing progress! With too much to read & absorb, papers in more crowded fields are citing new work less, and canonizing highly-cited articles more.
55
1K
4K
The wisdom of the crowd works even if the crowd is totally drunk. This study got undergrads intoxicated & found that while drunk individuals make a lot more errors, the consensus of groups of drunk people was as accurate as that of groups of sober people.
73
556
4K
Whatever bad faith arguments you see on Twitter today, Schopenhauer already thought of how to use them 115 years ago as part of his 38 Stratagems to unfairly win an argument. Twitter is powered by Stratagems 2, 3, and 22, along with a bit of Strategy 16.
67
816
4K
The new BloombergGPT AI may be harbinger of the next wave of corporate AI. Current AIs are trained on web data (though firms can add their own training). BloombergGPT is 52% either proprietary data or cleaned financial data. And it shows signs of being better at financial tasks.
58
610
4K
Three things about advanced voice on ChatGPT:.1) It is as good as the demo.2) It is clearly capable of producing a lot more audio (I occasionally get sound effects, etc) but there are guardrails.3) It is super weird. Lots of unconscious cues make it feel like talking to a person.
22
53
801
No matter the language, we exchange information at 39 bits/second, suggesting a biological limit. Languages that are lower information density are spoken fast (Spanish & Japanese) while denser languages are spoken more slowly (Thai, Vietnamese, Chinese).
56
1K
4K
Nice analogy. We are at the KT Boundary for information. Archivists should lock down the pre-2023 information world. What comes after is going to be… different.
@emollick Like radiocarbon dating pre/post nuclear era.
17
349
4K
Researchers left 17,000 wallets on the streets of 355 cities, some empty, some with money. Contrary to the predictions of economists, people everywhere were more likely to return wallets with money in them. But rates did vary from country to country.
114
1K
4K
The ability of multimodal AI to “understand” images is underrated. I just took these. Given the first photo Claude guesses where I am. Given the second it identifies the type of plane. These aren’t obvious.
90
251
4K
👀This paper argues that half of venture capital investments are “predictably bad” (causing 10% losses - or $900M out of $9B in investments) because VCs put too much emphasis on founder talent, and also aren’t good at identifying who is talented (MBAs are especially undervalued).
104
535
4K
If AGI is achievable, the very first people who will know it will be traders who will suddenly find all of their good strategies & trades don’t work anymore, and that an unknown firm is on the winning side of all of them.
238
232
4K
There are now papers showing that the migration of academic communities from Twitter to other social media sites mostly failed. But they never fully returned to Twitter either. I wonder if we will ever see so many different communities interacting in the same online space again.
130
860
4K
This is just science - go for a walk. This paper found walking (whether outdoors or a treadmill) increased key types of creative thinking for over 80% of undergraduates. The reasons are not fully clear, but there seem to be direct effects on the brain
Running may be better exercise than walking, but walking is better for having ideas. I do get ideas when running, but not nearly as many. Possibly running uses more of your brain, whereas walking uses very little, judging from the fact that people can do it while asleep.
53
688
4K
Before you get sick of OpenAI tricks, one fun one. You & an android are in front of a judge. The judge tells each of you to say one word. They will then kill whoever they think is the AI based on that. An MIT paper calls this the "minimum Turing test." What do you say? 1/
288
558
3K
If you want to learn how to learn, these two charts are where you should start. They are the result of a large meta-analysis of study techniques. You should skip the highlighting, summarizing, and rereading. Instead, practice, quiz & explain. Open paper:
36
729
3K
Neat! Hey, GPT-4: "Invent a board game using emoji that we can play against each other. Give it a theme from Shakespeare's Tempest. Set up the board and explain the rules & let us play each other.".🏝️🌊🌴🌴🌴🌊. 🌊🌴🏰🌴🌊🌊. 🌴🌴🌴🌊🌴🌴. 🌊🌊🌊🌴🌊🌊 .🌴🌴🌴🌊🌴🌴 .🌊🌊🌴🌴🌴🏝️
50
364
4K
In this study, AI was more accurate than two thirds of radiologists, yet when radiologists had AI help their diagnoses did not improve. Why?. Humans ignored the AI’s advice when it conflicted with their views. A big barrier to future human-AI collaboration
132
664
3K
Avoiding "ums" and "ahs" when you speak may actually make what you say less memorable, since these "speech disfluencies" appear to serve a real purpose. They boost a listener's memory of whatever comes immediately afterwards by focusing our attention.
45
597
3K
The thing about GPT-4 that most feels like a superpower is that I can work with it to write programs to automate annoying tasks. I have already co-created two Python programs in the last week to solve little problems, and a Unity program for fun. I don't know these languages.
67
291
3K
Management consultants, keep a close eye on AI. As an experiment, I fed GPT-4 the official practice McKinsey cases, including the math. It nails it, often better than the official answer. (As far as I can tell, these launched after the training data window for GPT-4 concluded)
75
477
3K
Dang, Google's veo 2, same prompt.
"golden retriever running through a brutalist art gallery with modern art on the walls as a single red balloon floats overhead" in Sora, Kling (the best Chinese model), Runway (best American model). Best of 2 videos. This is a hard prompt, so look at consistency, reflections, etc
88
232
3K
Forget losing jobs to AI, look to the sky!. Flocks of pigeons detect cancer in scans with 99% accuracy & “are suitable surrogates for human observers in certain medical image perception studies, thus avoiding the need to recruit, pay & retain clinicians. ”
88
576
3K