The Algorithmic Bridge 2023 in 60 Memorable Sentences

Tired: Summaries. Wired: Random excerpts without context

Merry Christmas friends!

I didn’t publish on Monday because I was spending Christmas with my family. Hope you had a great day, too, eating and drinking a lot — enough to justify another New Year's resolution to lift those weights that, like mine, are gathering dust in the basement.

I also hope you received the presents you deserved. Perhaps not GPT-4.5 (Sam is surely getting some coal this year) but at the very least another yearly TAB subscription…

Anyway, if you’re not sure if you want one for $50, what about $35? This generous offer brought to you by Santa expires on January 1st, 2024, so get yours now!

Anyway, to the important stuff. I’ve been preparing this article for a while. It’s something I’ve never seen anyone do (bad idea, you should always imitate), so it’s kinda risky.

I wanted to do a review with highlights of what we’ve covered on TAB in 2023. I thought of summarizing each article or enumerating my most popular posts but I feel the former is too boring and the latter too typical and not very informative (also, I don’t want to repeat what the other 2146512 AI newsletters are doing).

That’s why I’ve decided to re-read TAB’s 2023 archive in its entirety and have taken my favorite 60 sentences or small paragraphs (five per month), to share with you. I think this approach can work well because of six reasons:

• I can tease you to read the articles without spoiling the surprise.

• You have, in one place, the highest-quality excerpts curated by me.

• These are great prompts for you to write your own essays or reflect.

• It’s a concise overview for 14,254 of you who found TAB after Jan 1st, 2023.

• It’s a quick reminder for 6,384 of you who came before that.

• Most importantly: You can easily skim this by month to prevent boredom.

Let’s see how this experiment goes and let me know how you like it in the comments!

January

How AI Can Help With the Loneliness Epidemic (FREE)

A lot of people simply don’t enjoy the possibility to feel human connection whenever they please … for them, an AI companion isn’t the worse out of the two choices but … the only one they have.

ChatGPT and the Future (Present) We're Facing (PAID)

The education system has a lot of room for improvement. If it hasn’t changed in so many years it’s because there weren’t strong-enough incentives to do so. ChatGPT gives us a reason to reimagine education.

4 Optimistic Takes on the Future of AI (PAID)

AI winters are a reaction to unfulfilled expectations, not just busted bubbles — I believe that today we have enough fulfilled expectations to avoid bad weather.

AI Influencers From the Post-ChatGPT Era (FREE)

The only way AI had to go mainstream was through this path … We love clickbait. We love hype. We love easy content. We love shortcuts. We don't like hard stuff that requires energy, time, and effort. That’s why AI influencers even exist in the first place.

5 Practical Applications Where ChatGPT Shines (FREE)

If I found a magic wand, I wouldn’t be able to tell where its limits lie and would presuppose it’s capable of doing a lot of things it may not be able to … That’s ChatGPT in the minds of most people. A magic wand.

February

Why Generative AI Angers Artists but Not Writers (FREE)

What [artists] fear is that non-artists (i.e. the majority of art consumers) wouldn’t care [about AI’s mistakes] due to our inability to appreciate visual art to its finest detail. To artists, generative AI may be a bust but for the untrained rest of the world, it’s just fine.

Google's Bard May Prove Everyone Wrong (FREE)

Do [investors and shareholders] agree that language models + search may not be a good idea yet? … Google’s dilemma is between shipping a half-baked product — with all the consequences it implies — and not satisfying those who keep the company’s pockets full.

A New Way of Doing — and Experiencing — AI (FREE)

We’re going to see less open R&D and more production-focused efforts. A decade of advances fueled by good practices is coming to an end and will give way to a new era — during hard times, survival and competition often supersede caution, openness, safety, and cooperation.

Prompt Engineering Is Probably More Important Than You Think (FREE)

Prompt engineering will be, pretty much literally, the only means we’ll have to talk to these increasingly powerful aliens. And we don’t really know in which ways they’re aliens because we can’t directly ask — except with prompts.

Learn To Master Prompt Engineering With This Singular (Triple) Framework (PAID)

In some sense, artistic creativity reigns there where science and engineering fall short. When it’s not about precise understanding but about abstract intuition.

March

On the Dangers of Overused AI Metaphors (FREE)

When I read [the term “stochastic parrot”] now, I realize it doesn’t play the role it was conceived for anymore because its ideological charge impedes it; it’s no longer a pointer to some deep truth about the unreliable nature of language models but a loaded expression that signals the author’s partisanship.

3 Prompt Engineering Lessons You Must Know (PAID)

“Sampling can prove the presence of knowledge but not the absence” … You can always be one more (better optimized) interaction away from unveiling a new capability [from e.g. ChatGPT] you were unaware of.

GPT-4: The Bitterer Lesson (FREE)

Soon, we’ll be just spectators, mere observers of a world neither built by us nor understood by us. A world that unfolds before our eyes — too fast to keep up, and too complex to make sense.

How OpenAI Fooled Us All (FREE)

OpenAI isn’t the sole culprit [for everything that’s gone wrong in AI] but it was the one we trusted. The one that promised to be different. A non-profit to benefit all of humanity. And we fell for it. It fooled us all.

Schrödinger's AGI (PAID)

AGI is like Schrödinger's cat — its untameable blurriness will prevent us from knowing if it exists or not.

April

AI Can’t Benefit All of Humanity (FREE)

First, technology disrupts the poor more, and then, even if it eventually benefits everyone, it benefits the rich more … Even the infinite resources of [an AI-driven] post-scarcity world would be unevenly distributed under an imperfect system.

How to Remain Human in the Age of AI (PAID)

It doesn’t matter if we aren’t that special … We need to believe we are special. It is the belief and not the supposed truth it holds, that is at the core of being human.

ChatGPT: Eraser of the Implausible (FREE)

ChatGPT would be 100% right, 100% of the time if we were all the same person. The more distinct we are from each other, the more it misses to capture our character in all its essence.

No Writer Is Safe From AI (PAID)

At some point in the future, we will return to the numbers of pre-printing days; only a tiny percentage of people will … write … But we’ll normalize that loss, as we have done [with so many other things] before.

Size Doesn't Matter (PAID)

The vast majority of people … don’t care about having access to the best of the best. We care about the best quality-price relationship … It doesn’t matter if big is what grabs our attention. Small seems now the winning bet.

May

Why I (Don't Use ChatGPT to) Write (FREE)

There’s one main reason … why it wouldn’t make sense for me to use ChatGPT. It’s as simple as it gets: It would steal the pleasure I find in writing.

The Missing Piece in Geoffrey Hinton's Newfound Fear of AI Existential Risk (PAID)

[Geoffrey Hinton] isn’t afraid that AI will get smarter than us or become uncontrollable or kill us because he knows something we don’t. He’s afraid because he doesn’t know.

Why No One Wants AI Ethics Anymore (FREE)

Here’s my advice for [AI ethicists]: give up the aggressive discourse around the performance and “intelligence” of AI systems … Even if you’re mostly right, you can’t fight against the perception of millions of people.

OpenAI Rushed Up the Ladder It Is Now Kicking Away (PAID)

Being a newcomer, [Sam Altman] could secure a privileged position for OpenAI as the government's right-hand man in the industry and gain the ability to keep the space perhaps not unregulated, but underregulated, in a way that would benefit his company while erecting hurdles for potential competition.

Generative AI Is Not a Magic Wand (PAID)

Generative AI is many things — useful, interesting, entertaining, and even problematic — but it’s not a world-shaking revolution. Very few things in history have withstood the test of time to be called “revolutionary.” Generative AI doesn’t seem to be one of them.

June

A Non-Cynical Reading of AI Risk Letters (FREE)

[The superintelligence prophets] want to be the architects of a future that in its inevitability can only be embraced, in its intractability must capture all our efforts, and in its desirability should be brought about faster — all those things mixed together into a rather incomprehensible simultaneity.

ChatGPT Will *Always* Make Stuff Up — What Can We Do About It? (PAID)

Factuality is orthogonal to the design metrics used to train language models … the problem [of confabulation] doesn’t have a solution because nothing needs to be solved.

GPT-4's Secret Has Been Revealed (PAID)

If we’re nitpicky, GPT-4 is the result of having, on the one hand, enough money and GPUs to train and run eight ~GPT-3.5 models stacked together and, on the other, the audacity to dust an old technique invented by another company without telling anyone.

All ChatGPT Users Are AI Psychologists (PAID)

The crude attempts we’ve taken to craft … the correct way to think about [AI systems like ChatGPT] are hopeless by their contemporaneity to the birth of the very things they refer to.

How the Great AI Flood Could Kill the Internet (FREE)

The internet, once a vibrant agora of high-quality underground sites democratized and decentralized for creators and consumers, intended to satisfy the needs of those who wanted to escape the mainstream channels, will be deformed beyond recognition into a desert populated by AI bots.

July

3 Endings More Poetic Than AI Wiping Us Out (FREE)

Not everything that ends is death. The Big Crunch is reincarnation; the Big Freeze is a static forever. Like those, the endings I’ve described … aren’t about death, for not everything that ends needs to die.

Google DeepMind’s Response to ChatGPT Could Be the Most Important AI Breakthrough Ever (PAID)

DeepMind’s Alpha family and OpenAI’s GPT family each possess a secret sauce that has turbocharged progress toward human-level intelligence in recent years. This secret sauce mix is what I believe Google DeepMind plans to imbue into Gemini.

ChatGPT: A Bullshit Tool For Bullshit Jobs (FREE)

No one wants to be a bullshiter all day long yet many people have to. ChatGPT can’t help but be one. The match couldn’t be more perfect. ChatGPT isn't emptying their professional lives of meaning. No, it’s emptying them of the modern illness of meaninglessness.

How Smart Is AI, Really? (Or: Why xAI May Be Good News) (PAID)

That's exactly what AI has always needed but never achieved: Mathematics-based robust theoretical groundwork to sustain the otherwise brittle empirical effectiveness of deep learning.

If You Want to Master Generative AI, Ignore All (But Two) Tools (FREE)

Life is too short to be chasing all the time after things we don’t really need.

August

3 Powerful Strategies (Other Than AI Detectors) That Teachers Can Adopt to Adapt to Generative AI (FREE)

The only question [teachers] need to ask … Given that I’m forced by external circumstances to do it this way, what’s the route I should take to not only compensate for the loss but create a net win?

ChatGPT Is the Perfect Creative Machine — Except For One Big Little Thing (FREE)

ChatGPT is a thing because of the Shakespeares of the past who collectively contributed to humanity’s archive.

The GPU Shortage Has Forced AI Companies to Take a Break That Could Last Until 2024 (PAID)

The urgent issue currently keeping … LM builders … awake at night isn’t design-related: it is the global GPU shortage. There aren’t enough high-quality, high-performance chips to satisfy demand.

ChatGPT Killed the Old AI. Now Everyone Is Rushing to Build a New One (PAID)

The next algorithmic breakthrough is probably further today than it was a few years ago (before the ChatGPT boom).

AI Companies Have Exhausted Human-Made Data. They're Now Using AI to Make More (PAID)

Chips are made by chipmakers, algorithms are discovered by researchers, but data is created through a collective, non-directed, and constantly evolving effort that involves the whole of humanity.

September

ChatGPT's Remarkable Success Raised the Bar Too High for Everyone—Even OpenAI Itself (PAID)

OpenAI is the architect of an achievement of such caliber that it has redefined what we thought was possible — and will be endlessly applauded by some yet unendingly despised and condemned by others for it.

What No One Outside OpenAI Can Really Understand About OpenAI (FREE)

Radford and Sutskever. And Altman and Brockman. And the others. They did not only build GPT, GPT-2, GPT-3, GPT-4, and of course ChatGPT. They lived through it … It wasn’t the value of the achievement that set them apart but the intimateness with which they went through it.

Why AI Can't Make Human Creativity Obsolete (FREE)

Humans like humans.

Perhaps AI Is Modern Alchemy. And That’s Not a Bad Thing (FREE)

[AI researchers] are … the humblest of all when it comes to not ruling out avenues of exploration and investigation … that could eventually yield foundational insights for our future.

AGI Has Been Achieved Internally (FREE)

What will happen once someone, generally accepted as an authority in the space, asserts with a confidence above some threshold — and being in possession of some kind of evidence — that AGI has been achieved?

October

Economist Tyler Cowen Explains Why AI Won’t Be the Revolution We Expect (PAID)

AI has yet to meet the friction inherent to our very human world to prove, over decades-long continued growth, that it deserves to be called a revolution.

It’s Time We Move On From Chatbots (PAID)

Chatbots will soon be an obsolete way to think about frontier AI — the future is not about better chatbots, but about something completely different.

Appendix to The Techno-Optimist Manifesto (FREE)

Technology, progress, and productivity growth aren’t a yes-or-no question, as [Marc] Andreessen seems to portray them to be in the manifesto, but a yes-or-no plus a what/who/how question. For me, it’s a yes — then we will see.

Forget About AI Girlfriends. Here’s What the Future of Human-AI Relationships Looks Like (PAID)

[AI companions are] mirrors that confirm that [our] experience is, after all, valid. Mirrors that distract [us] from the terrible fact that [our] perceived world only lives in [our] minds.

The Most Important Skill in the 21st Century (FREE)

Today, all people’s troubles come from not having to know how to sit still in one room. It’s not that we don’t know how to be bored or in silence. It’s that we don’t need to know.

November

Sam Altman's Hidden Message at OpenAI DevDay Keynote (PAID)

The real AI state-of-the-art lives, for the rest of us, years in the future of our imagination.

Here's Why People Will Never Care About AI Risk (FREE)

People will never care about AI existential risk as much as say, climate change or nuclear war, because it’s intrinsically abstract to think about it.

Here's Why We Are Living During the Most Beautifully Balanced Era for Humans and AIs (PAID)

Chess grandmasters haven’t lost their love for the game but that doesn’t change that they know — and can’t not know — that no human will ever again be the best chess player in the universe.

10 Controversial Things I Believe About AI That I Shouldn't Say Out Loud (PAID)

It’s … important to not hide in the safety of abstract ideas, thought experiments, and endless debates and settle down on a specific attitude about things.

What You're Feeling Is Shiny Object Syndrome (FREE)

The tragedy of OpenAI’s success and its impeccable marketing: Make it look easy and people will assume it is.

December

One Year of ChatGPT: These Six Tendencies Have Been Quietly Shaping the Future of AI (PAID)

[ChatGPT] replaced people’s deluded illusions about the future of technology by giving them something to touch, to play with, something to experience.

How to Survive as a Human Creator in the AI Era (FREE)

Creators who nurture their relationships and remain as human as they can be are not in trouble in the AI era. However, those whose goal is empty, impersonal content-creating face a dilemma — not because AI will become more like them but because they are already a lot like AI.

AI Models Are Tragic Slaves of Their Sublime Predictive Accuracy (PAID)

Language models catastrophically fail to imagine what doesn’t exist — what shouldn’t exist … that makes them unable to perceive the past as we do. And worse, unable to discover a better future.

God Is Dead So They Are Building a New One (PAID)

I see what Schmidhuber perceives as the incoming bearer of unfathomable complexity; what Sutton seeks as our ascended successor; what Altman thinks will host our fragile digitalized minds; what LeCun believes we can keep controlled under chains; what Hinton is starting to be profoundly terrified of. A new god.

The Algorithmic Bridge 2023 in 60 Memorable Sentences (FREE)

Writing this article has taught me this lesson, which I share with you here in the form of a meta excerpt: There seems to be no correlation between the work that I value most or consider the highest quality and the work that gets the most attention or praise. Corollary: Keep writing.