A Robot has no Soul


Overthinking the Future

Episode V

Art and Artificial Intelligence

I hear a lot of people worrying about AI replacing art and creativity. It's undeniable that tools like Midjourney and ​Suno​ are marvelous works of engineering, but I don't think that we should be worried about the future of art.

AI superpowers don't make an artist

While today's AI tools have made it easy for anyone to create beautiful images and songs, I believe the best AI art is made by artists. Only someone who has spent years studying a topic – understanding the vocabulary, terminology, and the interplay between certain techniques and their emotional perception – can use a tool like Midjourney with precision and mastery. Which is why OpenAI and Google are partnering with artists to showcase their tools. This also explains why so many people feel frustrated when image generators don't create what they had in mind; they lack the right words to guide the AI effectively.

This idea applies to any field. Giving an average person AI superpowers won't make them extraordinary. Today's AI is still just a tool. If I gave you the most high-tech, super-powered chisel in the world, would you suddenly become a master carpenter? Probably not.

Think about Spider-Man. Getting bitten by a radioactive spider didn’t turn Peter Parker into a superhero overnight – in fact, that's the whole point of the story. Having the ability doesn't make you a superhero. Peter Parker needed months of practice and effort to master his newfound powers and exert intuitive and willful control over them.

Technology has disrupted art many times before

Most people today don't remember that cinema used to feature music played by live orchestras. When cinemas introduced digital audio in the late 1920s, it was met with massive resistance from the industry. Today, everyone is delighted to enjoy the marvels of spatial audio in the cinema, paying only $10 for a movie – something impossible when you had to pay an entire orchestra for every show.

This wasn't a one-off event.

In the 1980s, mixtapes made music production accessible to everyone, eliminating the need for expensive studios controlled by record labels. Then, in the 2000s, digital audio workstations (DAWs) took creative control to a whole new level, transforming home computers into studios-in-a-box.

And today, we have AI that can create entire songs just from text descriptions, giving unprecedented music creation power to the people.

When art is in danger, humanity will be as well

For me, worrying about AI replacing artists misses the forest for the trees. Today's AI is little more than statistics on steroids. Perhaps the biggest tell-tale is the inability of current AI models to generalize into tasks beyond their training data. The models don't have goals, consciousness, sentience, or emotions.

Creating an AI that poses an existential risk to human art has implications far beyond art itself. An AI that displays emotional depth and artistic vision is probably an AI that can replace not just artists, but also engineers, scientists, and politicians. When AI reaches that level, we will need to worry about the future of the entire humanity in a world where AI can do everything faster, better, and cheaper.

But that's a conversation for another time.

(I stole the title from this great post)

🌎 From the Internet

OpenAI, the company behind ChatGPT, is becoming increasingly harder to trust.

  • Back in November, the board tried to oust the CEO, concerned about his honesty. The CEO managed to return, removing the board instead. Ilya Sutskever, the board member and co-founder with the most technical expertise on AI, secluded himself from public attention for months after that. A few weeks ago, a former board member said that the board learned about the existence of ChatGPT from Twitter. Nice.
  • The company made employees sign an NDA prohibiting them from criticizing the company after they left, threatening their equity (which is the bulk of an employee’s salary at OpenAI). Employees weren’t informed about this NDA at the time of joining. The CEO claimed he didn’t know about it (let's just say this is extremely unlikely).
  • The company disbanded its Superalignment team, which focused on making AI safe and aligned with human values. Ilya Sutskever, who led it, left OpenAI. Jan Leike, who co-led it, also left, posting that OpenAI is not taking safety seriously on his way out.
  • Yesterday, OpenAI announced that the NSA's former Director, the person responsible for spying on everyone on the planet, will be joining the company’s board. Edward Snowden wasn't amused.

Sorry for the bad vibes, I try to keep the newsletter's content positive and uplifting, but OpenAI is giving me the creeps. Just be mindful about what you are sharing with ChatGPT.


twitter profile avatar
Balaji
Twitter Logo
@balajis
4:38 PM • Jun 16, 2024
438
Retweets
3280
Likes

Remember how I asked where the flying cars are in a previous episode of the newsletter? Here they are! China is in the process of approving regulations for flying cars, which is great because the western part of the hemisphere will follow. Finally, flying cars!


Udio is one of the coolest AI apps that I tried recently. You write a description, and it generates a 30-second music clip. Then, you can add more parts, including an intro and an outro, to create a full song! AND you can add custom lyrics in any language you want. If you ever wanted to write a song for your mom but didn't have the skill, now's your time to shine as the favorite child.

Tip: Use ChatGPT to generate lyrics.

📘 Using AI Language Tools Effectively

In the last episode I shared with you that I am writing a small handbook for using AI. I already shared a draft with you yesterday in a private email! Check yesterday's email to download it.

If you subscribed to the mailing list after the email, (a) welcome! and (b) don't worry – you will get an updated version very soon.

Feedback is already pouring in and I'm genuinely happy to hear that it's useful.

So, I read your LLM handbook and I have to say it's amazing! Honestly, very useful. All the terminology and explanations were very clear and easy for me to understand.
Great! The writing style, premise, and general format and length are excellent. I found myself smiling at the intro. I wonder what feedback you'll get from others but from me, I love it.

Playing with different "ads" is super fun. I'm thinking to run some A/B tests to see what works better.

About those spelling and grammar issues, thank you to everyone who found them so far. Keep them coming! You're making the next version better ❤️

Share the handbook with your friends: https://newsletter.aristot.io/llm-ebook

📚 Week's Reading Highlight

One last thing! I have a question for you:


Yes! I managed to send the episode within 14 days of the previous one. Time to go sleep. 🫡 I'm also on X (Twitter). I post there a lot.

Do you know someone who might like this newsletter? Share it with them to make two people happy: https://newsletter.aristot.io

-Aristotelis

P.S. What I'm doing now: https://aristot.io/now

Aristotelis' Private Mailing List

This is my personal mailing list. I share project updates and thoughts on topics including tech, science, and AI. In rarer occasions I am sharing my tips on note taking, AI tools, and productivity.

Read more from Aristotelis' Private Mailing List

Overthinking the Future Episode IV Why Google's AI Overviews being so horribly wrong is O.K. Ok, so, here's what happened. During their annual developer conference called I/O, Google released a new feature called "AI overviews." When you search on Google, apart from the regular search results, you also get an AI-generated summary at the top. This summary provides the gist of what you're looking for based on the search results. Some people even claimed that this feature will kill Perplexity,...

Overthinking the Future Episode III Is a world where superintelligence exists fun? I recently started thinking about superintelligence, and I'm not sure if I want us to build it. I'm still processing my thoughts, but it's a feeling that started growing as I thought about various future scenarios where superintelligence exists. This episode's main text ended up being a lot longer than I thought, so instead of making the email super long, I made it a full post on my website. It's the first...

Overthinking the Future Episode II Where's my flying car?! Calvin and Hobbes, December 1989 ©️ I'm reading Hall's "Where is my flying car?" and the introduction made me think a lot. If you compare planes in 1910 with 1960, they're on a totally different tier of technological magnificence. A plane from the 60s looks almost identical to one from 2010. Wright Model B, 1910 (left). Boeing 707, 1958 (middle). Boeing 787, 2011 (right). It may sound normal to us that we don't have flying cars yet,...