In today’s newsletter, Part four of our five-part seriesI will try to convince you that large language models are already good for a wide variety of tasks – and they are getting smarter every day.
In the last newsletter, my colleague Cade Metz wrote that AI Large Language Models (LLMs) can be unstable and unreliable—providing false information and behaving strangely towards users.
I’ve been using AI tools like ChatGPT almost daily for months and I’ve seen them give a lot of wrong answers. (And don’t remind me of my creepy encounters with Bing.)
But I’ve also seen these AI programs do amazing things — feats of creativity, flexibility, and efficiency that amaze me. I worry that in trying to curb the AI hype, skeptics are missing how groundbreaking — and potentially disruptive — the technology is.
Large language models write poems and screenplays.One of the latest GPT-4s achieved the 90th percentile on the bar exam, and got the highest score On many Advanced Placement exams. The potential productivity gains for workers are enormous. (exist a study, programmers who use GitHub Copilot (LL.M. for coders) complete tasks 56% faster than programmers who don’t. )
Here are six things AI is particularly good at right now:
Explain concepts at multiple difficulty levels
Over the past few months, whenever I’ve been curious about a new topic, I’ve used ChatGPT as my personal tutor. Let’s say, if it’s fractional reserve banking, I’ll start by asking ChatGPT to “explain fractional reserve banking at the high school level.” (Or, if I’m still confused, “in middle school.”)
ChatGPT and other AI chatbots can handle these types of prompts well, leveraging their skills natural language processing, as long as you don’t ask about recent events or extremely obscure topics. Visiting a tutor with infinite patience and zero cost to guide me in new subjects felt like a superpower. When used correctly, they can be amazing teaching tools.
Editing and Constructive Criticism
I still write my columns (including this newsletter) myself. But in the past few months, I have recruited ChatGPT as my assistant.
When I get stuck, I often post a few words to see if it sparks any ideas. If I want to strengthen an argument, I ask it to poke holes in my reasoning.
I’ve found Bing to be great for editing because it can find things on the Internet. (One of my favorite writing techniques is to have it read Strunk and White’s “Style Elements” And put forward revision suggestions according to the principles of this book. )
A new generation of chatbots
A brave new world. A new crop of chatbots powered by artificial intelligence has sparked a debate over whether the technology can upend the internet’s economy, rendering today’s powerhouses obsolete and creating the industry’s next giants. Here’s what to know about bots:
You can use a similar trick to use AI to improve your emails, cover letters, or any other writing you’re working on. As a reminder, you should always double check the AI’s work.
get out of trouble creatively
AI can also be a great tool to keep your creative juices flowing. Recently, I was trying to come up with questions to ask my podcast guests. I paste a guest’s resume into ChatGPT and ask it to give me “10 thoughtful, sharp interview questions” for this person. Of the issues it produced, most were fine and I ended up using a few on the show.
Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, suggests using AI to overcome writing blocks, or to get tough projects off the ground. For example, someone who wants to open a bakery can ask ChatGPT for a list of all the necessary equipment and permits.
“It’s easy to get distracted by a daunting challenge,” Mr Morik wrote in an essay. Recent Blog Posts. “AI offers a way to power yourself.”
Rehearse for real-world tasks
I’ve also been using ChatGPT and other AI apps to rehearse offline tasks that I find unpleasant or difficult.
When I have to have a difficult conversation with a friend, I engage ChatGPT in a role-playing exercise. “Pretend you’re my friend and react the way you think my friend might react,” I told it. Then I had a mock version of the conversation. It didn’t make the actual conversation easy, but I felt more confident knowing I’d practiced with the chatbot.
Of course, AI chatbots cannot replace human friendship. But they can be a kind of on-demand sounding board, giving us essential feedback and advice without judgment.
Everything about SparkNotes
One of the most powerful capabilities of AI language models is the ability to quickly summarize large amounts of text.
I use AI programs to “read” long texts for me and summarize dense academic papers. Often, even if some details are missing, they get it roughly right.
This capability really comes into play when you combine it with other AI capabilities such as transcribing audio and video recordings.Suddenly, it was possible to condense a two-hour podcast into bullet points, or get a concise summary of a work meeting while you’re still in a meeting.
If you get tired of reading this newsletter, you can ask ChatGPT or Bing to summarize the rest for you, or rewrite it as a snappy limerick. I promise I won’t be offended.
code even if you don’t code
People are already writing code with AI, whether they know any programming languages or not.
These models are not designed for writing software.But when they were trained on vast amounts of Internet text, including content from coding sites like GitHub and Stack Overflow, they learned how to code—a phenomenon known as emergency behavior. Now, if you type something like “build me a chrome extension that translates text from any website into a pirated language” you might get your code back an app that actually works.
Social media has been filled with examples of this over the past few months surprised man Who is building with AI tools Websites and Apps Faster than anyone – at a fraction of the cost.
What’s next?
All of this raises some very obvious questions about the future, such as: If the LLM is helping non-writers write well, non-coders write well, what other kinds of power are they unlocking? How many jobs will these new AI capabilities eliminate?
Tomorrow, Cade will be back to wrap up the series and see where all this development might lead.
your homework
What other chatbot skills can you discover?
Think of a task that takes up a lot of your daily mental energy, like a chore at work, or making a weekly dinner plan to feed your family. Then work with the chatbot to see how it performs.
(Last week, The Times interviewed people who use ChatGPT to have more productive conversations with teens, make up imaginative bedtime stories, and draft love letters to their spouses. The Times, which I co-host The podcast Hard Fork features stories from listeners who are using artificial intelligence to negotiate rent, understand medical outcomes, and affirm their gender identity.)
As always, share your best, worst, and funniest results in the comments.
test
Question 1 of 3
How is Kevin’s experience using artificial intelligence tools such as ChatGPT?
Select your answer to start the quiz.
Glossary
Natural Language Processing: Large language models are techniques used to understand and generate human language, including text classification and sentiment analysis. These methods often use a combination of machine learning algorithms, statistical models, and linguistic rules.
Emergency Act: Unexpected or unexpected capabilities in large language models, enabled by the model’s learning patterns and rules from its training data. For example, an LLM trained on programming and coding sites can write new code. Other examples include creative abilities such as composing poetry, music and fictional stories.
Click here for more glossary terms.
