User Overview

Followers and Following

Followers
Following
Trendsmap

History

Total Followers - Last Year
Daily Follower Change - Last Year
Daily Tweets - Last Year

Tweet Stats

Analysed 20,347 tweets, tweets from the last 684 weeks.
Tweets Day of Week (UTC)
Tweets Hour of Day (UTC)
Key:
Tweets
Retweets
Quotes
Replies
Tweets Day and Hour Heatmap (UTC)

Tweets

Last 50 tweets from @emollick
In reply to @emollick
Another Zoom meeting with Sauron. pic.twitter.com/DrhQ18zCuy
Replying to @emollick
"Good meeting. Want to get lunch? Who's in Alice, Todd, Khamûl the Easterling, Bill?"
 
Anxiety about the future is the price for humanity’s superpower.

Humans alone can engage in complex prospection, thinking about futures. Unlike chimps, we plan for mutual incompatible futures, at the cost of constant worry about what will happen. Paper: adambulley.org/wp-content/upl…
 
In reply to @emollick
And as silly as it is, it also shows how AI can do a great job as a teaching tool by giving learners interesting connections For example, your eyes might glaze over if I try to teach you about why org charts exist, but maybe it would be better in Star Wars or the Sopranos? pic.twitter.com/g6QeQ1zWul
Replying to @emollick
Another Zoom meeting with Sauron.
 
In reply to @emollick
"Today is a good day to scrum." pic.twitter.com/LCBZZVAHdr
 
This is fun: Bing does process re-engineering for fictional evil organizations.

You are a management consultant, analyze the org chart of Mordor. How would you improve it to make it more flexible in winning in today's changing Middle Earth? Provide KPIs.

Write a re-org memo. pic.twitter.com/cWB2MHMyVk
Replying to @emollick
And as silly as it is, it also shows how AI can do a great job as a teaching tool by giving learners interesting connections For example, your eyes might glaze over if I try to teach you about why org charts exist, but maybe it would be better in Star Wars or the Sopranos?
 
One way AI is going to change research is by making hard and inaccurate data analysis much easier.

Here GPT-3.5 (not even GPT-4!) outperforms human annotators and costs 20x less per annotation. (Data annotation is one of the most annoying & expensive parts of research projects)
Replying to @emollick
Any special issues on GPT methods coming out soon?
 
One way AI is going to change research is by making hard and inaccurate data analysis much easier.

Here GPT-3.5 (not even GPT-4!) outperforms human annotators and costs 20x less per annotation. (Data annotation is one of the most annoying & expensive parts of research projects)
LLMs Can Outperform Humans on Data Annotation

-Compare 0-shot accuracy of ChatGPT vs crowd-workers on:
relevance
topic detection
stance detection
general frame detection
policy frame detection

-ChatGPT better on 4/5 tasks
-20x cheaper than MTurk

Paper: arxiv.org/abs/2303.15056 pic.twitter.com/VS3wRCkuUG
 
In reply to @emollick
I tried entering this into Bing chat but didn't work. can I ask what you're using to access this feature?
Replying to @Peter_1_1
Put it into creative mode?
 
In reply to @emollick
I tried using the prompts in your post to see if I can generate good examples for learning concepts in 1st year physics. The results were... underwhelming... it might have been my prompts. what are some things you did to optimize output?
Replying to @Peter_1_1
Tell me what didn't work! Did you use GPT-4?
 
This is fun: Bing does process re-engineering for fictional evil organizations.

You are a management consultant, analyze the org chart of Mordor. How would you improve it to make it more flexible in winning in today's changing Middle Earth? Provide KPIs.

Write a re-org memo.
 
I had Bing develop a bestiary of a planet, and the story of how each creature was discovered. This is what it made

(Hint: To get Bing's AI-led images to work well, be in creative mode & tell Bing "Create a vivid description and pass that to Bing Image Creator to make a picture") pic.twitter.com/0idl0kuexg
Replying to @emollick
I could see this being an interesting school project where students have to work together with Bing to create a realistic ecosystem. Initial prompt: We are on an alien planet like Mars a billion years ago, and you are my computer guide and tutor. You are going to produce a… twitter.com/i/web/status/1…
 
I had Bing develop a bestiary of a planet, and the story of how each creature was discovered. This is what it made

(Hint: To get Bing's AI-led images to work well, be in creative mode & tell Bing "Create a vivid description and pass that to Bing Image Creator to make a picture")
 
In reply to @emollick
How do you get 1 in 12? The figure says 750k, the total US population was 62M, which if half the pop was men, and about 4million male children, is 1/35 working in railroads. So I must be missing something.
Replying to @OneShoup @maiamindel and 1 otherfalse
Next page, from the ICC
 
As a reminder, asking ChatGPT for quotes or passages from books will result in hallucinations.

The couple of quotes I checked here were false. They may (or may not) represent the spirit of the book, but Chat is not connected to the internet & doesn’t retrieve exact text for you. pic.twitter.com/GiDQbKt1FS
Replying to @emollick
A lot of “magic prompts” from Twitter aren’t that magical. Try asking Bing or GPT-4 what you want to accomplish and ask it to guide you. It will help you do some perfectly decent prompting. And try to accidentally avoid forcing it to lie. oneusefulthing.substack.com/p/how-to-get-a…
How to Get an AI to Lie to You in Three Simple Steps
oneusefulthing.substack.com
 
As a reminder, asking ChatGPT for quotes or passages from books will result in hallucinations.

The couple of quotes I checked here were false. They may (or may not) represent the spirit of the book, but Chat is not connected to the internet & doesn’t retrieve exact text for you.
 
We have had waves of technological change that restructured the entire economy before.

In 1890, 1 out of every 12 men in America was employed in railways, which used 56% of all horsepower across all machines. And as @calebwatney pointed out, nails alone were .5% of the 1810 GDP! pic.twitter.com/ssEp3gdv72
Replying to @emollick
Usually automation shifts folks into better jobs, but not always: When ☎️ operators were rapidly replaced (15% of all women had worked as operators at some point!). Young folks quickly found other jobs, veteran operators had a lifelong hit to wages. nber.org/papers/w28061
 
We have had waves of technological change that restructured the entire economy before.

In 1890, 1 out of every 12 men in America was employed in railways, which used 56% of all horsepower across all machines. And as @calebwatney pointed out, nails alone were .5% of the 1810 GDP!
 
We build workflows around our tools. Agile development, sales & marketing pipelines, and even organizational form👇 are enabled by & work along with our technology

I see folks adding GPT to existing workflows. That is temporary. Ultimately, work is going to be reorganized for AI
Replying to @emollick
There are already documents made by AI being sent to people who have AI read then and respond. There are whole categories of (hopefully mostly annoying) coordination and reporting work that can and must be redone because they were built for humans. It will take time, of course.
 
We build workflows around our tools. Agile development, sales & marketing pipelines, and even organizational form👇 are enabled by & work along with our technology

I see folks adding GPT to existing workflows. That is temporary. Ultimately, work is going to be reorganized for AI
The first org chart was made for the New York & Erie Railroad in 1854, solving issues of control & coordination in ways we still use. Not only is it full of plant-inspired beauty, it also combines a map of the physical railroad & one of company structure. mckinsey.com/business-funct… pic.twitter.com/Sqh6670AFW
 
Bing, quick: the mission is in jeopardy!

I need to assemble a crack team of agents to enter Doctor Madman's lair and rescue Agent X, and stop him before his Explode-o-Beam destroys the world. Who is on my team? Give me a picture of them ready to deploy... pic.twitter.com/JFrKkSiaEE
Replying to @emollick
There are two real implications of this goofy spy drama: 1. You can see a future where AI might create games on demand that could be very personalized & compelling. 2. It is easy to push AI into a personality you suggest & think you are talking to a real person with real goals
 
Bing, quick: the mission is in jeopardy!

I need to assemble a crack team of agents to enter Doctor Madman's lair and rescue Agent X, and stop him before his Explode-o-Beam destroys the world. Who is on my team? Give me a picture of them ready to deploy...
 
In reply to @emollick
And here is the most horrifying Simpsons ever, along with Calvin and Hobbes, Avatar the Last Airbender, and Batman pic.twitter.com/GoxBqSmzO4
Replying to @emollick
Having Bing describe, without using specifics, and then create paintings based on the prompts: Guernica, Mona Lisa, Persistence of Memory, Degas's In a Cafe...
 
Most people use AI art tools for just a few subjects.

Here's an interactive map of over 6.5M generated images, mapped by similarity. I pointed out a few key areas.

At the center of all generated art lies classical statues of Snoop Dogg & anime characters atlas.nomic.ai/map/809ef16a-5… twitter.com/i/web/status/1…
 
In reply to @emollick
And I should note that neither the temperature or token dials work yet. Future project. And the only real change I made was switching it to using GPT-3.5 (It only knows about GPT-3)
Replying to @emollick
Also added a few more buttons: Devils Advocate: Give me the argument against this point Solutions: Give me 10 solutions to this problem Connections: Who might be good experts on this topic (a little hallucinatory until I can hook GPT up to the internet)
 
This paper finds folks are less happy when they let their mind wander: “a human mind is a wandering mind, and a wandering mind is an unhappy mind. The ability to think about what is not happening is a cognitive achievement that comes at an emotional cost.” renevanmaarsseveen.nl/wp-content/upl… pic.twitter.com/mk7z73VD5Z
Replying to @emollick
Part of the reason is our secrets. 97% of people have a secret they are keeping & the average person has 13! "Frequency of mind-wandering to, but not concealing, secrets predicts lower well-being.” Here are the secrets of a random sample of New Yorkers. columbia.edu/~ms4992/Pubs/2…
 
If you are skeptical about the impact of AI on real, day-to-day work, you may be interested in my little experiment, where I got as far as I could in marketing a new product launch in 30 minutes.

AI did hours of work in minutes, with high-quality drafts. oneusefulthing.substack.com/p/superhuman-w…
 
This paper finds folks are less happy when they let their mind wander: “a human mind is a wandering mind, and a wandering mind is an unhappy mind. The ability to think about what is not happening is a cognitive achievement that comes at an emotional cost.” renevanmaarsseveen.nl/wp-content/upl…
 
In reply to @mattsclancy
Pretty sure @emollick is run by an AI now so are we really surprised the AI wants us talking more about it?!?
Replying to @orgRem @mattsclancy
THAT. DOES. NOT. COMPUTE
 
In reply to @emollick
Isn't this just DALLE-2 in Bing?
 
New game with Bing's drawing feature.

I ask it to describe to its "Bing Image Creator" classical paintings in as much detail as possible, but not using the names of the artist or the style, and see what comes out. Here is Monet, van Gogh, Rothko, Vermeer.

Interesting results. pic.twitter.com/JD1rZA6u59
Replying to @emollick
And here is the most horrifying Simpsons ever, along with Calvin and Hobbes, Avatar the Last Airbender, and Batman
 
New game with Bing's drawing feature.

I ask it to describe to its "Bing Image Creator" classical paintings in as much detail as possible, but not using the names of the artist or the style, and see what comes out. Here is Monet, van Gogh, Rothko, Vermeer.

Interesting results.
 
In reply to @emollick
I tried heat transfer and asked it to explain temperature and conduction to 7th graders. It used words like temp is average kinetic energy, which I don't think a 7th grader knows. I asked it to try again w/ simpler words and it went too far so Temp was now just heat energy.
Replying to @mckennagene
I also see you are using 3.5 - try Bing creative mode or GPT-4
 
Replying to @mckennagene
Math is its achilles heel. But not for long.
 
In reply to @emollick
"I want to make it so that when I hit a button on my stream deck, it sends whatever I just copied to GPT to comment on, and shows the results on the screen." It suggests installing OBS "I don't want to use OBS" There were 2 minor bugs (I pasted in error text, it fixed them)
Replying to @emollick
And I should note that neither the temperature or token dials work yet. Future project. And the only real change I made was switching it to using GPT-3.5 (It only knows about GPT-3)
 
One cool thing about AI is that it lowers the friction of little projects. I wanted to create hotkeys so that I could push one and GPT would take the text I copied and explain it, translate it, riff on it, etc

No plan & I don't know Python. Just asked ChatGPT. Took 10 minutes. pic.twitter.com/ZMi15pb8JT
Replying to @emollick
"I want to make it so that when I hit a button on my stream deck, it sends whatever I just copied to GPT to comment on, and shows the results on the screen." It suggests installing OBS "I don't want to use OBS" There were 2 minor bugs (I pasted in error text, it fixed them)
 
One cool thing about AI is that it lowers the friction of little projects. I wanted to create hotkeys so that I could push one and GPT would take the text I copied and explain it, translate it, riff on it, etc

No plan & I don't know Python. Just asked ChatGPT. Took 10 minutes.
 
In reply to @emollick
If the link doesn’t work, paste this into ChatGPT: You generate clear, accurate examples for students of concepts. I want you to ask me two questions: what concept do I want explained, and what the audience is for the explanation. Provide a clear, multiple paragraph explanation… twitter.com/i/web/status/1…
Replying to @emollick
Also, as far as I can tell, Bing only really does GPT-4 level output in creative mode. (Something happened to balance mode to make it work much worse)
 
In reply to @emollick
I can try creative mode, but I think the link you shared defaulted to some other mode. I'm getting a lot of stuff like this (I prompted for an illustrative example). pic.twitter.com/DHm9WBK9oD
Replying to @mpershan
Yuck. Links used to work better. Try using the general purpose prompt in the second tweet.
 
In reply to @tomcoates
Annoying. The copy paste only works in Edge browsers I think, not in the Bing app. Summoning @MParakhin to see if that is right, since I think embedded tutors is a really exciting concept.
Replying to @emollick @tomcoates and 1 otherfalse
See the second tweet in the thread for the prompt to use.
 
Let's try a new educational experiment on Twitter. To explain it, you are going to need to deal with a (short) long Twitter post. But, basically, rather than doing a thread to explain some complex concepts, I have set up Bing AI to act as a tutor.

Why? I oneusefulthing.substack.com/p/using-ai-to-…twitter.com/i/web/status/1…
Replying to @emollick
If the link doesn’t work, paste this into ChatGPT: You generate clear, accurate examples for students of concepts. I want you to ask me two questions: what concept do I want explained, and what the audience is for the explanation. Provide a clear, multiple paragraph explanation… twitter.com/i/web/status/1…
 
In reply to @emollick
This didn't work very well for me, I can't seem to get bing to tell me anything about Diffusion of Innovation other than stuff like "innovators are characterized by those who want to be the first to try the innovation."
Replying to @mpershan
Frustrating. Even in creative mode?
 
In reply to @emollick
I can’t get it to open any of these links on mobile.
Replying to @tomcoates
Annoying. The copy paste only works in Edge browsers I think, not in the Bing app. Summoning @MParakhin to see if that is right, since I think embedded tutors is a really exciting concept.
 
Let's try a new educational experiment on Twitter. To explain it, you are going to need to deal with a (short) long Twitter post. But, basically, rather than doing a thread to explain some complex concepts, I have set up Bing AI to act as a tutor.

Why? I oneusefulthing.substack.com/p/using-ai-to-…twitter.com/i/web/status/1…
Using AI to make teaching easier & more impactful
oneusefulthing.substack.com
 
Whether you look at teams of scientists, inventors, or software developers, this paper shows that small teams are more likely to develop disruptive, novel ideas than large teams. Fascinatingly, people who do breakthroughs in small groups can't in big ones. nature.com/articles/d4158…
 
Papers show humans accept AI more if we have an illusion of control.

Reminds me of the story of how the space shuttle was designed for automatic landings but military pilots insisted that they be in charge. So, lowering landing gear (but nothing else) required flipping a switch.
 
Ethan Mollick Retweeted ·  
In reply to @emollick
Too much focus on “What Does AI Mean” & “Can We Create Super-intelligence.” Those are important, but speculation. Too little focus on how AI changes the balance between firms and employees, how it affects the nature of jobs, how we will deal with bad actors, how to regulate it.
 
Once AI starts to generate movies and games, the personalized media future will be wild.

"Bing, I am Ethan Mollick. Create a picture to delight me, and tell me why you think it would delight me specifically" (it gets the name of the game I made wrong, but everything else right)
 
In reply to @TheAaronBowley
Not intended! I genuinely want to know whether you get better results by asking AI to pump things up beyond the maximum scale. Like if I want a movie poster, is telling it “150% Fresh Rotten Tomatoes” better than 100%?
Ha! That was a prompt someone shared (no longer remember where). Prompt crafting used to be ritual more than actuality, in any case. You didn't know what worked, you just said the words.
 
Ethan Mollick Retweeted ·  
Professor @emollick goes ham on AI.

In just over 30min, he had AI: do market research, create a positioning doc, write an email campaign, create a website, logo, hero image, make a social media campaign, and script and create a video.

In 30 minutes.

oneusefulthing.substack.com/p/superhuman-w…
 
I don’t think “prompt crafting” is going to be a key skill of the future & I say that as someone who is good at prompting AI. Two things happening:
1) AI systems are getting much better at guessing what you want without help
2) AI firms are incentivized to make prompting easier
Replying to @emollick
For a preview look at Midjourney v5 (image with white cup): Madeleine and a cup of warm tea Midjourney v4: “Madeleine and a cup of warm tea , Professional Food Photography, Five Michelin Star, High Class, Elegant, Delicious, Delectable, Savory, Alluring, Cinematic, Color… twitter.com/i/web/status/1…
 
 
For access to this functionality a Trendsmap Explore subscription is required.

A Trendsmap Explore subscription provides full access to all available timeframes

Find out more

Thanks for trying our Trendsmap Pro demo.

For continued access, and to utliise the full functionality available, you'll need to subscribe to a Trendsmap Pro subscription.

Find out more

This account is already logged in to Trendsmap.
Your subscription allows access for one user. If you require access for more users, you can create additional subscriptions.
Please Contact us if you are interested in discussing discounts for 3+ users for your organisation, or have any other queries.