Quotes

[On using generative AI for work despite the risk of errors:] AI is helpful despite being error-prone if it is faster to verify the output than it is to do the work yourself. For example, if you're using it to find a product that matches a given set of specifications, verification may be a lot faster than search. There are many uses where errors don't matter, like using it to enhance creativity by suggesting or critiquing ideas. * At a meta level, if you use AI without a plan and simply turn to AI tools when you feel like it, then you're unlikely to be able to think through risks and mitigations. It is better to identify concrete ways to integrate AI into your workflows, with known benefits and risks, that you can employ repeatedly.

[…] in the era where these AI systems are true 'everything machines', people will out-compete one another by being increasingly bold and agentic (pun intended!) in how they use these systems, rather than in developing specific technical skills to interface with the systems. We should all intuitively understand that none of this will be fair. Curiosity and the mindset of being curious and trying a lot of stuff is neither evenly distributed or generally nurtured. Therefore, I'm coming around to the idea that one of the greatest risks lying ahead of us will be the social disruptions that arrive when the new winners of the AI revolution are made - and the winners will be those people who have exercised a whole bunch of curiosity with the AI systems available to them.

"...in case of need I will make bombards, mortars, and firethrowing engines of beautiful and practical design." It starts with Archimedes focusing the sun’s rays upon the fleet at Syracuse, it starts with the first rock hurled by the first grasping hand. It starts where we start. It is mind, it is hunger, it is greed, it is defense, it is mischief, it is the devil, it is the god; it is life...

I think you overstate the usefulness of the [jQuery Rules] plugin. Using this plugin, users are now limited by what selectors that can use (they can only use what the browsers provide - and are at the mercy of the cross-browser bugs that are there) which is a huge problem. Not to mention that it encourages the un-separation of markup/css/js.

It is a sound maxim, and one which all close thinkers have felt, but which no one before Bentham ever so consistently applied, that error lurks in generalities: that the human mind is not capable of embracing a complex whole, until it has surveyed and catalogued the parts of which that whole is made up; that abstractions are not realities per se, but an abridged mode of expressing facts, and that the only practical mode of dealing with them is to trace them back to the facts (whether of experience or of consciousness) of which they are the expression. Proceeding on this principle, Bentham makes short work with the ordinary modes of moral and political reasoning. These, it appeared to him, when hunted to their source, for the most part terminated in phrases. In politics, liberty, social order, constitution, law of nature, social compact, &c., were the catch-words: ethics had its analogous ones. Such were the arguments on which the gravest questions of morality and policy were made to turn; not reasons, but allusions to reasons; sacramental expressions, by which a summary appeal was made to some general sentiment of mankind, or to some maxim in familiar use, which might be true or not, but the limitations of which no one had ever critically examined.

At GitHub, we want to protect developer privacy, and we find cookie banners quite irritating, so we decided to look for a solution. After a brief search, we found one: just don’t use any non-essential cookies. Pretty simple, really. 🤔 So, we have removed all non-essential cookies from GitHub, and visiting our website does not send any information to third-party analytics services.

We believe that AI tools are at their best when they incorporate and represent the full diversity and breadth of human intelligence and experience. [...] Because copyright today covers virtually every sort of human expression– including blog posts, photographs, forum posts, scraps of software code, and government documents–it would be impossible to train today’s leading AI models without using copyrighted materials. Limiting training data to public domain books and drawings created more than a century ago might yield an interesting experiment, but would not provide AI systems that meet the needs of today’s citizens.

The Great Man … is colder, harder, less hesitating, and without respect and without the fear of “opinion”; he lacks the virtues that accompany respect and “respectability”, and altogether everything that is the “virtue of the herd”. If he cannot lead, he goes alone. … He knows he is incommunicable: he finds it tasteless to be familiar. … When not speaking to himself, he wears a mask. There is a solitude within him that is inaccessible to praise or blame.

Also note that the python visualizer tool has been basically written by vibe-coding. I know more about analog filters -- and that's not saying much -- than I do about python. It started out as my typical "google and do the monkey-see-monkey-do" kind of programming, but then I cut out the middle-man -- me -- and just used Google Antigravity to do the audio sample visualizer.

The blog post announcing the shutdown was done one day early. The idea was to take the opportunity of the new Pope being announced and Andy Rubin being replaced as head of Android, so that the [Google] Reader news may be drowned out. PR didn't apparently realize that the kinds of people that care about the other two events (especially the Pope) are not the same kind of people that care about Reader, so it didn't work.

Private blockchains are completely uninteresting. (By this, I mean systems that use the blockchain data structure but don't have the above three elements.) In general, they have some external limitation on who can interact with the blockchain and its features. These are not anything new; they're distributed append-only data structures with a list of individuals authorized to add to it. Consensus protocols have been studied in distributed systems for more than 60 years. Append-only data structures have been similarly well covered. They're blockchains in name only, and -- as far as I can tell -- the only reason to operate one is to ride on the blockchain hype.

The problem is a lack of respect for the consumer. The manufacturers don't act as if the computer belongs to you. They act as if it is a billboard for restricted trial versions of software and ads for Web sites and services that they can sell to third-party companies who want you to buy these products.

Allo shows the ultimate failure of Google's Minimum Viable Product strategy. MVP works when you have almost no competition, or if you are taking a radically different approach to what's on the market, but it completely falls on its face when you are just straight-up cloning an established competitor. There's no reason to use a half-baked WhatsApp clone when regular WhatsApp exists.

There is a common-place book argument, Which glibly glides from every tongue; When any dare a new light to present, "If you are right, then everybody's wrong"! Suppose the converse of this precedent So often urged, so loudly and so long; "If you are wrong, then everybody's right"! Was ever everybody yet so quite?

The boring yet crucial secret behind good system prompts is test-driven development. You don't write down a system prompt and find ways to test it. You write down tests and find a system prompt that passes them. For system prompt (SP) development you: - Write a test set of messages where the model fails, i.e. where the default behavior isn't what you want - Find an SP that causes those tests to pass - Find messages the SP is missaplied to and fix the SP - Expand your test set & repeat

Last September, I received an offer from Sam Altman, who wanted to hire me to voice the current ChatGPT 4.0 system. He told me that he felt that by my voicing the system, I could bridge the gap between tech companies and creatives and help consumers to feel comfortable with the seismic shift concerning humans and AI. He said he felt that my voice would be comforting to people. After much consideration and for personal reasons, I declined the offer.

We're at a critical juncture in the evolution of software. The web is still here and it is still strong. Anyone can still put any information or applications on a web server without asking for permission, and anyone in the world can still access it just by typing a URL. I don't think I appreciated how important that is until recently. Nobody designs new systems like that anymore, or at least few of them succeed. What an incredible stroke of luck the web was, and what a shame it would be to let that freedom slip away.

Two bicycles side by side are passing on the flower path– the fragrance stuck in the spokes falls on the road and glitters ... Hey there, Hey there, I called, but I couldn’t even face their eyes directly Having called them to stop, I couldn’t give them any words, because they seemed to know more beautiful roads than I know

There are contexts in which it is immoral to use generative AI. For example, if you are a judge responsible for grounding a decision in law, you cannot rest that on an approximation of previous cases unknown to you. You want an AI system that helps you retrieve specific, well-documented cases, not one that confabulates fictional cases. You need to ensure you procure the right kind of AI for a task, and the right kind is determined in part by the essentialness of human responsibility.

My goal is to keep SQLite relevant and viable through the year 2050. That's a long time from now. If I knew that standard SQL was not going to change any between now and then, I'd go ahead and make non-standard extensions that allowed for FROM-clause-first queries, as that seems like a useful extension. The problem is that standard SQL will not remain static. Probably some future version of "standard SQL" will support some kind of FROM-clause-first query format. I need to ensure that whatever SQLite supports will be compatible with the standard, whenever it drops. And the only way to do that is to support nothing until after the standard appears. When will that happen? A month? A year? Ten years? Who knows. I'll probably take my cue from PostgreSQL. If PostgreSQL adds support for FROM-clause-first queries, then I'll do the same with SQLite, copying the PostgreSQL syntax. Until then, I'm afraid you are stuck with only traditional SELECT-first queries in SQLite.

This [technological] pattern has happened over and over and over again. Someone built the system, they assumed certain user behaviors. The users came on and exhibited different behaviors. And the people running the system discovered to their horror that the technological and social issues could not in fact be decoupled.

Yeah, unfortunately vision prompting has been a tough nut to crack. We've found it's very challenging to improve Claude's actual "vision" through just text prompts, but we can of course improve its reasoning and thought process once it extracts info from an image. In general, I think vision is still in its early days, although 3.5 Sonnet is noticeably better than older models.

Want to know if your ‘HTML application’ is part of the web? Link me into it. Not just link me to it; link me into it. Not just to the black-box frontpage. Link me to a piece of content. Show me that it can be crawled, show me that we can draw strands of silk between the resources presented in your app. That is the web: The beautiful interconnection of navigable content

After every war someone has to clean up. Things won’t straighten themselves up, after all. Someone has to push the rubble to the side of the road, so the corpse-filled wagons can pass. ...Those who knew what was going on here must make way for those who know little. And less than little. And finally as little as nothing. In the grass that has overgrown causes and effects, someone must be stretched out blade of grass in his mouth gazing at the clouds.

Students who use AI as a crutch don’t learn anything. It prevents them from thinking. Instead, using AI as co-intelligence is important because it increases your capabilities and also keeps you in the loop. […] AI does so many things that we need to set guardrails on what we don’t want to give up. It’s a very weird, general-purpose technology, which means it will affect all kinds of things, and we’ll have to adjust socially.

To see what is in front of one’s nose needs a constant struggle. One thing that helps toward it is to keep a diary, or, at any rate, to keep some kind of record of one’s opinions about important events. Otherwise, when some particularly absurd belief is exploded by events, one may simply forget that one ever held it. Political predictions are usually wrong. But even when one makes a correct one, to discover why one was right can be very illuminating. In general, one is only right when either wish or fear coincides with reality. If one recognizes this, one cannot, of course, get rid of one’s subjective feelings, but one can to some extent insulate them from one’s thinking and make predictions cold-bloodedly, by the book of arithmetic. In private life most people are fairly realistic. When one is making out one’s weekly budget, two and two invariably make four. Politics, on the other hand, is a sort of sub-atomic or non-Euclidean word where it is quite easy for the part to be greater than the whole or for two objects to be in the same place simultaneously. Hence the contradictions and absurdities I have chronicled above, all finally traceable to a secret belief that one’s political opinions, unlike the weekly budget, will not have to be tested against solid reality.