Quotes

And there never was a time, I believe, when those who read at all, read so many more books by living authors than books by dead authors; there never was a time so completely parochial, so shut off from the past...Individualistic democracy has come to high tide: and it is more difficult today to be an individual than it ever was before.

OAuth came out of my worry that if the Twitter API became popular, we'd be spreading passwords all around the web. OAuth took longer to finish than it took for the Twitter API to become popular, and as a result many Twitter users' passwords are scattered pretty carelessly around the web. This is a terrible situation, and one we as responsible web developers should work to prevent.

If Web authors actually use this feature, and if IE doesn't keep losing market share, then eventually this will cause serious problems for IE's competitors — instead of just having to contend with reverse-engineering IE's quirks mode and making the specs compatible with IE's standards mode, the other browser vendors are going to have to reverse engineer every major IE browser version, and end up implementing these same bug modes themselves.

Sam Altman expelling Toner with the pretext of an inoffensive page in a paper no one read would have given him a temporary majority with which to appoint a replacement director, and then further replacement directors. These directors would, naturally, agree with Sam Altman, and he would have a full, perpetual board majority - the board, which is the only oversight on the OA CEO. Obviously, as an extremely experienced VC and CEO, he knew all this and how many votes he (thought he) had on the board, and the board members knew this as well - which is why they had been unable to agree on replacement board members all this time.

I cannot accept your canon that we are to judge Pope and King unlike other men, with a favourable presumption that they did no wrong. If there is any presumption it is the other way, against the holders of power, increasing as the power increases. Historic responsibility has to make up for the want of legal responsibility. Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men, even when they exercise influence and not authority, still more when you superadd the tendency or the certainty of corruption by authority. There is no worse heresy than that the office sanctifies the holder of it…You would hang a man of no position like Ravaillac; but if what one hears is true, then Elizabeth asked the jailer to murder Mary, and William III of England ordered his Scots minister to extirpate a clan. Here are the greatest names coupled with the greatest crimes; you would spare those criminals, for some mysterious reason. I would hang them higher than Haman.

When I worked at Amazon.com we had a deeply-ingrained hatred for all of the SQL databases in our systems. Now, we knew perfectly well how to scale them through partitioning and other means. But making them highly available was another matter. Replication and failover give you basic reliability, but it's very limited and inflexible compared to a real distributed datastore with master-master replication, partition tolerance, consensus and/or eventual consistency, or other availability-oriented features.

I am once again shocked at how much better image retrieval performance you can get if you embed highly opinionated summaries of an image, a summary that came out of a visual language model, than using CLIP embeddings themselves. If you tell the LLM that the summary is going to be embedded and used to do search downstream. I had one system go from 28% recall at 5 using CLIP to 75% recall at 5 using an LLM summary.

A complex system that works is invariably found to have evolved from a simple system that worked. The inverse proposition also appears to be true: A complex system designed from scratch never works and cannot be made to work. You have to start over, beginning with a working simple system.

God is dead. God remains dead. And we have killed him. How shall we comfort ourselves, the murderers of all murderers? What was holiest and mightiest of all that the world has yet owned has bled to death under our knives: who will wipe this blood off us? What water is there for us to clean ourselves? What festivals of atonement, what sacred games shall we have to invent? Is not the greatness of this deed too great for us? Must we ourselves not become gods simply to appear worthy of it? ...This prodigious event is still on its way, still wandering; it has not yet reached the ears of men. Lightning and thunder require time, the light of the stars requires time, deeds, though done, still require time to be seen and heard. This deed is still more distant from them than the most distant stars—and yet they have done it themselves.

Scaling laws allow us to precisely predict some coarse-but-useful measures of how capable future models will be as we scale them up along three dimensions: the amount of data they are fed, their size (measured in parameters), and the amount of computation used to train them (measured in FLOPs). [...] Our ability to make this kind of precise prediction is unusual in the history of software and unusual even in the history of modern AI research. It is also a powerful tool for driving investment since it allows R&D teams to propose model-training projects costing many millions of dollars, with reasonable confidence that these projects will succeed at producing economically valuable systems.

It feels like we’re at a bit of an inflection point for the Django community. [...] One of the places someone could have the most impact is by serving on the DSF Board. Like the community at large, the DSF is at a transition point: we’re outgrowing the “small nonprofit” status, and have the opportunity to really expand our ambition and reach. In all likelihood, the decisions the Board makes over the next year or two will define our direction and strategy for the next decade.

GPT-4, like GPT-3 before it, has a capability overhang; at the time of release, neither OpenAI or its various deployment partners have a clue as to the true extent of GPT-4's capability surface - that's something that we'll get to collectively discover in the coming years. This also means we don't know the full extent of plausible misuses or harms.

If you have worked in search, you know how freaking hard even getting started with something close to this with traditional methods. Now, you can zero-shot it. > System Instructions: As a query categorization expert, you try to break down the intent of a search query. First, provide your reasoning and then describe the intent using a single category (broad, detailed, comparision) > > User: The query from the user is "nike versus adidas for terrain running". The user is a female, age 22. > > Model: The user is clearly looking to compare two specific brands, Nike and Adidas, for a particular activity, terrain running. While the user's demographics might be helpful in some situations (e.g., recommending specific product lines), the core intent remains a comparison. Category: Comparison There's a lot of hand-waving around query intent classification; it's always been like that. Now, it's straightforward (add a few examples to improve accuracy). But my point is that you could only dream about building something like this without having access to lots of interaction data.

Bill Gates has pulled off one of the greatest hacks in technology and business history, by turning Microsoft's success into a force for social responsibility. Imagine imposing a tax on every corporation in the developed world, collecting $100 per white-collar worker per year, and then directing one third of the proceeds to curing AIDS and malaria.

At this point all I could honestly tell you from the point of view of the editor of several of the HTML5 documents being held up is that the W3C have said they're won't publish without the objections being resolved, and that the objection is from Adobe. I can't even tell what I could do to resolve the objection. It seems to be entirely a process-based objection.

Spotify introduced the vocabulary of missions, tribes, squads, guilds, and chapter leads for describing its way of working. It gave the illusion it had created something worthy of needing to learn unusual word choices. However, if we remove the unnecessary synonyms from the ideas, the Spotify model is revealed as a collection of cross-functional teams with too much autonomy and a poor management structure.

Today we’re excited to launch ARC-AGI-2 to challenge the new frontier. ARC-AGI-2 is even harder for AI (in particular, AI reasoning systems), while maintaining the same relative ease for humans. Pure LLMs score 0% on ARC-AGI-2, and public AI reasoning systems achieve only single-digit percentage scores. In contrast, every task in ARC-AGI-2 has been solved by at least 2 humans in under 2 attempts. [...] All other AI benchmarks focus on superhuman capabilities or specialized knowledge by testing "PhD++" skills. ARC-AGI is the only benchmark that takes the opposite design choice – by focusing on tasks that are relatively easy for humans, yet hard, or impossible, for AI, we shine a spotlight on capability gaps that do not spontaneously emerge from "scaling up".

Third, X fails to provide access to its public data to researchers in line with the conditions set out in the DSA. In particular, X prohibits eligible researchers from independently accessing its public data, such as by scraping, as stated in its terms of service. In addition, X's process to grant eligible researchers access to its application programming interface (API) appears to dissuade researchers from carrying out their research projects or leave them with no other choice than to pay disproportionally high fees.

...In the heart of that light, lucid and inevitable, all that was scattered cohered. Superbright and all its progeny stood plain before him in conception and in detail and in its component part and its deepest strategies and in its awful and enticing radiance. He saw the design and the making of that device complete, and of further devices without end, and he stood apart from them as if it mattered not at all whether the deviser was himself or whether they came into being sooner or later. Trembling he stared across the burning fields and whispered, —Stop. Stop. But the traffic rushed on.

There was a time when you could whip out a parser in lex and yacc, stitch together a naive VM and throw it over the wall and you'd have a new scripting language. Those days are coming to a close and in a few years (if not months) you won't be able get traction with anything unless it does direct threading, is register based, has generational GC, does peephole optimizations, does trace-folding, does type-inferenced inline caching, etc.

Graphic designers had a similar sea change ~20-25 years ago. Flyers, restaurant menus, wedding invitations, price lists... That sort of thing was bread and butter work for most designers. Then desktop publishing happened and a large fraction of designers lost their main source of income as the work shifted to computer assisted unskilled labor. The field still thrives today, but that simple work is gone forever.

Craigslist is fighting back. Its latest gimmick is phone verification. Posting in some categories now requires a callback phone call, with a password sent to the user either by voice or as an SMS message. [...] Spammers tried using their own free ringtone sites to get many users to accept the Craigslist verification call, then type in the password from the voice message. Craigslist hasn't countered that trick yet.

Alles Grosse und Gescheite existiert in der Minoritaet. Es ist nie daran zu denken, dass die Vernunft populaer werde. Leidenschaft und Gefuehle moegen populaer werden, aber die Vernunft wird immer nur im Besitze einzelner Vorzueglicher sein. [“Every big and clever exists in the minority. You will never think of getting it majority. Passion and feelings can get popular, but sanity will only be the proper of some people.]

Imagine if 10% of the apps on iPhone came from Flash. If that was the case, then ensuring Flash didn’t break release to release would be a big deal, much bigger than any other compatibility issues. [...] Letting any of these secondary runtimes develop a significant base of applications in the store risks putting Apple in a position where the company that controls that runtime can cause delays in Apple’s release schedule, or worse, demand specific engineering decisions from Apple, under the threat of withholding the information necessary to keep their runtime working.

The way you motivate someone who doesn’t need the money is the same way you should motivate people who do need the money: by giving them meaningful roles with real responsibility where they can see how their efforts contribute to a larger whole, giving them an appropriate amount of ownership over their work and input into decisions that involve that work, providing useful feedback, recognizing their contributions, helping them feel they’re making progress toward things that matter to them, and — importantly — not doing things that de-motivate people (like yelling or constantly shifting goals or generally being a jerk).

Whenever you're called on to make up your mind, and you're hampered by not having any, the best way to solve the dilemma, you'll find, is simply by spinning a penny. No—not so that chance shall decide the affair while you're passively standing there moping; but the moment the penny is up in the air, you suddenly know what you're hoping.

Contributions must not include content generated by large language models or other probabilistic tools, including but not limited to Copilot or ChatGPT. This policy covers code, documentation, pull requests, issues, comments, and any other contributions to the Servo project. [...] Our rationale is as follows: Maintainer burden: Reviewers depend on contributors to write and test their code before submitting it. We have found that these tools make it easy to generate large amounts of plausible-looking code that the contributor does not understand, is often untested, and does not function properly. This is a drain on the (already limited) time and energy of our reviewers. Correctness and security: Even when code generated by AI tools does seem to function, there is no guarantee that it is correct, and no indication of what security implications it may have. A web browser engine is built to run in hostile execution environments, so all code must take into account potential security issues. Contributors play a large role in considering these issues when creating contributions, something that we cannot trust an AI tool to do. Copyright issues: [...] Ethical issues:: [...] These are harms that we do not want to perpetuate, even if only indirectly.

I went down into Shropshire to look at that famous library [of the late Lord Acton] before it was removed to Cambridge. There were shelves on shelves of books on every conceivable subject—Renaissance Sorcery, the Fueros of Aragon, Scholastic Philosophy, the Growth of the French Navy, American Exploration, Church Councils. The owner had read them all, and many of them were full in their margins with cross-references in pencil. There pigeon-holed desks and cabinets with literally thousands of compartments, into each of which were sorted little white slips with references to some particular topic, so drawn up (so far as I could see) that no one but the compiler could easily make out the drift of the section. I turned over one or two from curiosity—one was on early instances of a sympathetic feeling for animals, from Ulysses’ old dog in Homer downward. Another seemed to be devoted to a collection of hard words about stepmothers in all national literatures, a third seemed to be about tribal totems. Arranged in the center of the room was a sort of block or altar composed entirely of unopened parcels of new books from continental publishers. All had arrived since Lord Acton’s health began to break up. These volumes were apparently coming in at the rate of ten or so per week, and the purchaser had evidently intended to keep pace with the accumulation, to read them all, and to work their results into his vast thesis—whatever it was. For years, apparently, he had been endeavoring to keep up with everything that was being written—a Sisyphean task. Over all there were brown Holland sheets, a thin coating of dust, the moths dancing in the pale September sun. There was a faint aroma of mustiness, proceeding from thousands of 17th & 18th-century books in a room that had been locked up since the owner’s death. I never saw a sight that more impressed on me the vanity of human life and learning.

Find a level of abstraction that works for what you need to do. When you have trouble there, look beneath that abstraction. You won’t be seeing how things really work, you’ll be seeing a lower-level abstraction that could be helpful. Sometimes what you need will be an abstraction one level up. Is your Python loop too slow? Perhaps you need a C loop. Or perhaps you need numpy array operations. You (probably) don’t need to learn C.

Round, round, go round, Waterwheel, go round Go round, and call Mr. Sun Go round, and call Mr. Sun Birds, bugs, beasts, grass, trees, flowers Bring spring and summer, fall and winter Bring spring and summer, fall and winter ...Teach me how to feel If I hear that you pine for me, I will return to you.

Unfortunately, I was shocked, horrified and moderately surprised to see that nowhere is there any mention of how to encode negative numbers. Google, I appreciate you trying to help, and I understand that this grew out of needs for Google Finance, where stock prices can never dip below zero. But there's really not that much data out there in the real world that always exists solely above the origin.

"They are absolutely reasonable. I think that is their distinguishing characteristic. Yes, Mr. Erskine, an absolutely reasonable people. I assure you there is no nonsense about the Americans." "How dreadful!" cried Lord Henry. "I can stand brute force, but brute reason is quite unbearable. There is something unfair about its use. It is hitting below the intellect."

The language issues are indicative of the bigger problem facing the AI Pin, ChatGPT, and frankly, every other AI product out there: you can’t see how it works, so it’s impossible to figure out how to use it. [...] our phones are constant feedback machines — colored buttons telling us what to tap, instant activity every time we touch or pinch or scroll. You can see your options and what happens when you pick one. With AI, you don’t get any of that. Using the AI Pin feels like wishing on a star: you just close your eyes and hope for the best. Most of the time, nothing happens.