GPT-3, DALL-E, and our Multimodal Future [YouTube Video Series]

I like very much the key ideas in the #2 and #3 videos! Very strong and real arguments

1 Like

Fantastic insights, thank you so much for sitting down and chatting with me :heart:

2 Likes

All of this is good and all, you’re making it sound like this tech will be available to everyone at the same time, I don’t think so. DALL-E, codex , gpt4 ect. will be given to the Elite, or those who can afford it. Multimodal if for sure the future, but who gets it and when is the question. Just like codex is not available to everyone now (Which does not make sense). Pointless to have a tech, thats only 0.001% of the population has access to it.

This is a really good point. However, the 2-10 year timeline I’m talking about is when I speculate ordinary people will get access to it. You’re right, they may need to be a part of something like the OpenAI beta, but at least it will be available to ordinary people (maybe the beta will be much bigger by then). The series doesn’t attempt to guess when the ultra-elite, very privately funded, may create it independently and keep it all for themselves … that’s way too hard to know without inside information.

Pleasure was all mine!!! I want to turn the tables and interview you next on my podcast, I’ll DM you on Twitter about it in a few weeks.

1 Like

If everyone has access to the same AI tools, how will we come up with new things? Where will novelty come from?

I answer this really important question in today’s video:

Enjoy!

2 Likes

I appreciate your views on the multimodal approach, through my own personal research as a programmer it seems like this is the future. A smart friend of mine pitched me an idea for pretty much this exact concept (he doesn’t know ai implementation much he’s just smart) and that’s what sparked my research and then shortly after, google mentioned how they’re working on their perceiver which focuses on being multimodal and more general from a data input standpoint. I believe that people are starting to understand more what approaches we need in order to abstract the processes needed for human-like data interpretation and compression. What a time to be alive, for rich people at the top of companies lol.

1 Like

I agree with you @tyler.csurilla I’m very excited and grateful for the era we are living in. I’m also optimistic that many of the commercial challenges like latency, availability, and character limits will be addressed overtime.

1 Like

I enjoyed this video, it raised a lot of good questions.

I thought I’d address the concept of taste, and whether it can be taught. It can. Or rather, its development can be encouraged. Here’s how.

I taught graduate level creative writing for eight years. One of the most important parts was developing an individual reading list with my students. I’d have them come in with ten titles of books they liked. I’d then look at this list and come up with an assigned reading list that meshed with and elevated their taste. Say, for instance, that a student was a science fiction fan. I’d introduce them to European writers like Stanislaw Lem, Bruno Schulz, Italo Calvino–authors whose work was fantastical and otherworldly, but who paid closer attention to the beauty of language itself. The point was always to get them to demand more out of the art they consume. Once you’re able to spot the tropes in what you’re reading, you derive less pleasure from it, and you start seeking out more challenging work and elevating your taste. You can still honor the foundations of your taste while pushing it to become more sophisticated.

Taste is about not only defining what you like, but why you like it. There’s a quite excellent book in the 33 1/3 series of books on music about Celine Dion that is the best thing I’ve ever read about taste. Totally not joking. Here it is: https://www.amazon.com/Celine-Dions-Lets-Talk-About/dp/082642788X

This is an awesome insight @boudinot thank you so much for sharing. I’ve added the book to my personal reading list. This is great evidence that perhaps taste can be taught in schools (or at least elevated) like you mentioned, by spotting tropes and seeking greater influences, pushing your own work forward.

The kind of taste I’m proposing, I suppose, after rewatching my own video lol is the kind that becomes a dominant communication style, or something singular which enters the public zeitgeist in a really meaningful way. I suppose when I’m talking about “taste” I am collectively referring to taste, advancement, but also “setting” some kind of mainstream trend the masses adopt which becomes a norm … really influencing culture. I’m not saying what you’re proposing isn’t the pathway to that, but is there is something distinctive about this kind of taste (the kind that a society adopts) in your experience? Or are many of those cases elevated taste like you’re describing, with a one-of-a-kind life story that just resonates with society and the times?

There’s so much nuisance here of what makes great taste and what doesn’t. I think, as a society, we can spot poor taste in some shared way pretty easily, even though all of our tastes should be some kind of a subjective thing. One of my favourite subreddits is called Awful Taste but Great Execution, for the most part, as a collective we can agree many of things posted are in poor taste, but we may differ on our views on “great taste”. Many of my favourite things I did not like right away, it took some time for me to, “get over it” as they may have been something completely new. I guess I’d also love to ask you, what can we learn from poor taste? How does it help us define great/elevated taste?

Video #6 is up, folks!

It’s about the importance of art theory and design language, especially with Multimodal AI models in the future:

2 Likes

Good questions. I used to be super engaged with questions of taste, now I just like what I like and don’t theorize about it too much. I think you’re asking about how taste develops collectively. I don’t know how much I can offer on that, but I will say that it’s going to be really interesting when human writers are influenced by literature composed by AIs. It’ll be a really fascinating feedback loop.

2 Likes

I agree. I have a video coming up (I believe even this week) on AI based feedback.

Btw - the book is incredible so far, I’m on chapter 2/3. I never even thought about how much taste is driven by people’s own backgrounds and can be viewed relatively depending upon the audience. Thanks for sharing, I’m looking forward to finishing it hopefully this week.

Today’s video is a more philosophical one. It questions the “role” of human creatives in the future if AI can do a lot more of the work.

This video asks an important question and I’m excited to hear any thoughts and comments you all may have below.

2 Likes

Today’s video is more GPT-3 focused! I talk about using AI tools to get feedback on your work more often and how this trend will only accelerate in the future as AI models improve.

3 Likes

Today’s video is on this idea of “realtime creativity” and what that means for the future of creativity:

I’m really biased but honestly, this video is really important if you’ve tuned into any of the videos for the series so far.

2 Likes

I love the concept of “realtime creativity”. Effort does not equaly creativity. Further, not having the technical skills, or time for the effort doesn’t equal a lack of creativity. Anyhow, love this idea!

1 Like

I have a lot to say about this one.

Some context: I’m a writer who just spent ten years writing a novel. Producing creative work is constant tension between wanting the product to be finished and the time it actually takes to finish it. This is what you mean by “painstaking.” I have experienced that pain over and over and over again.

Having said that, I am convinced that there are no shortcuts in art. Works of art take the time that they take. I have been destroyed time and time again by rushing something that wasn’t ready. And in the process, I find that it is in the very quality of being painstaking where art’s greatest rewards lie. I would not trade the ten years it took me to write that novel for anything. If someone were to offer me a way to cut that time in half, I never, never, never would take that offer. It’s like removing the cumbersome and laborious sex part from the process of making babies.

What is far more exciting to me about incorporating AI into the creative process is applying the same level of painstaking attention with a new set of tools. I don’t see GPT-3 as a shortcut at all.

All the blind alleys, frustrations, deleted chapters, and tangents are just as much a part of the novel as the parts that readers will read. And these parts, that only the writer experiences, are never a waste of time.

The ego is seduced by shortcuts that promise external validation. We want people to like us, to post nice things about us on social media, to write glowing reviews, but these are never the true reward. The true reward is in the actual act, in the moment of creation, when it’s just you and the blank canvas or page. A positive review gives me a dopamine boost for a couple minutes. Knowing a character and witnessing their transformation over a decade is several orders of magnitude more rewarding.

3 Likes

I would say my video argues that creativity will indeed be realtime but our attention will then gives focus on experimentation and exploration which may take just as long. Imagine initializing the image with AI but spending 2-4 more hours experimenting and editing on top of what was already generated … getting it “just right”.

1 Like

Today’s video in the series is a GPT-3 written poem:

2 Likes