getpost

Yes, this is the context, but the study doesn't mention the Pauling protocol.

Pauling advocated doses much larger than 1 g. I'm not defending Pauling's position. I just want to point out that the study didn't examine Pauling's recommendation.

It so happens I have a copy of Vitamin C and the Common Cold (c) 1970. In Chapter 10 (page 86 of my paperback), in the section, "How to Ameliorate a Cold," he says, "At the first sign that a cold is developing, the first feeling of scratchiness of the throat, or presence of mucus in the nose, or muscle pain or general malaise, begin the treatment by swallowing one or two 500-mg tablets. Continue the treatment for several hours by taking an additional tablet or two tablets every hour."

"If the symptoms disappear quickly after the first or second dose of ascorbic acid, you may feel safe returning to your usual regimen. If, however, the symptoms are present on the second day, the regimen should be continued, with the ingestion of 4g to 10 g of ascorbic acid per day."

I often see this phenomenon when mainstream medicine designs studies to test claims by supplement advocates. It's hard to avoid the conclusion that many of these studies are intended to produce a negative result. I can't find the particular citation I have in mind at the moment, but there was a study on vitamin D for some illness, but the amount of vitamin D given was something like 400iu, the RDA (or maybe it was 800iu). The study then concluded that that dosage made no difference, which is probably is true, but idiotic. I mean, it's worth knowing that the low dose made no difference, but it doesn't justify the headline, "Vitamin D not helpful for ...." Also, I think the study used D2 (the plant form), which is not equivalent to D3, the animal form. (D2 provides some benefit if D3 is not available, but it is not the same thing.)

Point taken. I have a Python program that is an elemental version of awk, and I use that for the odd task. I can modify it if needed and I have the entire Python library to help me. Is the text Unicode? HTML? These little details matter.

I'm not complaining that someone banged out awk (speaking figuratively) on a Friday afternoon to do something and not have to stay after work. Excellent! My complaint is that the failure to address technical debt has negatively affected the productivity of millions, if not tens of millions, of people, often working under pressure, for DECADES.

I know lots of people like awk, but I pretend it doesn't exist. Why? Here's my comment on this from 6 years ago[0],

>I used awk until I learned Python (long ago). For me, awk was yet another example of the "worse is better" approach to things so common in unix. For example, if you make a syntax error, you might get a message like "glob: exec error," rather than an informative message. "Worse is better" is probably a good strategy in business and for getting things done, but still, mediocrity and the sense of entitlement that so often goes with carelessness, sickens me.

[0] https://news.ycombinator.com/item?id=13457265

Long live the Unix Hater's Handbook! (Unix is fine, and so are the criticisms herein. Some of these criticisms have been eclipsed by ongoing development.) https://en.wikipedia.org/wiki/The_UNIX-HATERS_Handbook

My experience is that drugs, including prescription anti-depressants, when they work, make it easier to see that your thinking is more arbitrary than you realize. Having been seemingly locked-in to a particular perspective, taking a drug, and seeing things a different way, breaks the lock. Then you can't go back to your old thought patterns, not if you're honest with yourself. This effect is independent of the particular effects that occur while the drug is active in your system.

Agreed; there should be no manipulation, but evidently that's too high a bar for most people most of the time.

Argh! The fact that false memories can be induced does not mean that all memories are false.

Commentary: https://www.newyorker.com/magazine/2021/04/05/how-elizabeth-...

Rebuttal: https://whyevolutionistrue.com/2021/04/25/the-new-yorkers-hi...

Nice and concise! I just started Proof and the Art of Mathematics by JD Hamkins[0], based on pg's recommendation[1], "is a beautiful book in both senses. It's both beautifully written, but also physically beautiful, thanks to its many illustrations, which I was surprised to hear were made by the author himself." Chapter 4 on induction improved my attitude.

[0]https://www.amazon.com/Proof-Art-Mathematics-Examples-Extens... [1] https://twitter.com/paulg/status/1662065331727155202

This phenomenon is inevitable, at least for any apps or websites that have to earn their keep. The problem is, features need to be added to expand the user base and keep up with, or stay ahead of, the competition. Adding almost any feature will make the interface more complex.

It's true that 95% of the users don't use the more complex features, but product reviews and corporate purchase decisions are driven by the feature set.

This reminds me of one of may favorite recipe books, Sacred and Herbal Healing Beers: The Secrets of Ancient Fermentation by Stephen Harrod Buhner[0], also published in 1998. It's much more than a recipe book, there's lots of history and information on herbalism.

[0] https://www.amazon.com/Sacred-Herbal-Healing-Beers-Fermentat...

The file is not in tabular format; that’s why I wanted Bard help, to process a highly irregular file. I was hoping to produce a tabular format.

I found an exciting feature—a way to submit a large amount of text—larger than you can paste in the Bard dialog window. (It's possible this isn't a new feature. Bard explained it to me this evening.) You can submit links to files in Google Drive. The links have to be publicly accessible. I just pasted the link to my file in Bard chat.

Bard can access the contents of the 322K file I pasted the link to. It definitely knows about the content of the file. I never said what it was about, but Bard knew it was about butterflies. It knew about content at the beginning of the file, and at the end.

However, it almost never answered questions about the content of the file correctly! For example, I asked it the number of species listed in the file and it said 109. There are 249 numbered species and some that are not numbered. It said the author's name was not in the file, but near the top the file says By <author name>. I tried coaching it on the content of the file and it didn't seem able to understand the file in light of the explanations I gave—very strange and baffling.

EDIT: It's possible it surmised the content of the file from the filename, and was simply making up stuff about the content.

Early this evening, I asked Bard if was updated to PaLM 2, and it said it was. I then asked it to write some Python programs, giving it more or less the same prompts I've given GPT4. Bard doesn't seem to be any better than it was a couple weeks ago in the cases I tried, and nowhere near as capable at GPT4. And it goes off the rails quickly. After even a short dialog (~5 statements), it becomes less and less able to stay on track and make coherent corrections to the code.

As the sqlite schema below indicates, each command has a unique id and a timestamp. Whether you want duplicates removed depends on what you want to know. It might be nice for the UI to expose a time context, which would retain duplicates. (Maybe it does! By coincidence, I just installed this yesterday, and I hardly know anything.)

CREATE TABLE history ( id text primary key, timestamp integer not null, duration integer not null, exit integer not null, command text not null, cwd text not null, session text not null, hostname text not null, deleted_at integer,

unique(timestamp, cwd, command) ); CREATE INDEX idx_history_timestamp on history(timestamp); CREATE INDEX idx_history_command on history(command); CREATE INDEX idx_history_command_timestamp on history( command, timestamp );

You can use sqlite[0] to export the database, or if you want a ui, use datasette[1]. On my mac, the database is stored at ~/.local/share/atuin/history.db

[0] https://sqlite.org/cli.html#export_to_csv [1] https://docs.datasette.io/en/stable/csv_export.html

There is an ongoing debate about the causes of obesity and, consequently, the right strategy for losing weight. There is a new paper out last week[0],[1]. "The surprising conclusion is we spend less energy when resting now than individuals did 30-40 years ago! The magnitude of the effect is sufficient to explain the obesity epidemic." The hypothesis is that the reduction in metabolic rate is to due to the relative decrease in the consumption of saturated fat!

[0] https://twitter.com/JohnSpeakman4/status/1651244232789598211 [1] https://www.nature.com/articles/s42255-023-00782-2.epdf?shar...

EDIT: Explainer video https://www.youtube.com/watch?v=nNNxE-K06KA

A certain webhost requires me to Authenticator for 2FA, and I did so. I also configured my iPhone to delete unused apps. Authenticator was unused, and got deleted, so naturally, I had to open a support ticket with the webhost to remove 2FA to regain access to my webhost account. I hope this feature will maintain my setup if the app gets deleted.

"Modernizing its search engine has become an obsession at Google"....

No doubt it can be modernized, but just making Google search as good as it used to be would be helpful. And why not better than it used to be with the traditional interface? That would still be worth doing even if Google adds on a superb AI.

Search is the proverbial goose that laid the golden egg and Google is killing it, and I don't mean "killing" in the contemporary ironic usage.

> constant cellular connection

This isn't at all required for music or maps. Phones have up to a terabyte of storage now.

The context length includes all of GPT4's replies, is that correct? If I want to minimize the context length, should I be asking GPT4 to limit its replies to just the fix that is requested?

GPT4 seems to prefer long-winded replies, i.e., when I ask for what amounts to a one-line fix, it repeats the entire block of code with the one line correction. In contrast, Replit's Ghostwriter often gives a concise reply showing only the one-line fix, and I have to ask for the entire block when it isn't clear where the fix is applied.

Hahaha, somehow I convinced myself that the rationale for the segment register scheme was a long term plan to accommodate virtual memory and memory protection. The idea being that you would only have to validate constraints on access when a segment register was loaded, rather than every single memory access.

Wow, this doesn't "just work" for me. I don't have an Apple Music subscription, but I've "purchased" plenty of classical music using iTunes, going back, what, 15+ years? The app doesn't appear to be designed for this use case. Why not? Isn't it just an elaborate skin on the basic Music app?

The app knows about my classical purchases, in that they are accessible from certain built-in playlists. For example, the piano pieces I own show up in Instrument Playlists/Piano. When I do a search, my music shows up among the search results, but lots of other results I don't own are also shown, so searching isn't a practical way to access my library.

None of my stuff shows up in Browse or Library. I have to "Add Albums" to my library? Just what I want, a learning curve, and extra effort, to play music I already own.

It appears Apple's main intention is to lock-in classical fans into a monthly subscription. Fine, but then just say, "This app requires a subscription to Apple Music."

There will always be new programs that matter; they matter because they are new.

Wait a minute, didn't I learn in college decades ago that deciding what is true is equivalent to the halting problem? Meaning, it can't be done in principle. There might be practical ways to ascertain the truth for a useful subset of statements in certain domains, but we are a very long way from knowing the truth about arbitrary statements, if that is even possible.

I didn't say it did. I posted the reply that I got, which did slightly improve my motivation to read the article.

It's interesting that you got a different response from Bard! I didn't alter the Bard response that I posted. Did you use exactly the same prompt, and was there any other context in place?

After the conversation I posted above, Bard and I discussed intractability vs undecideability, and it was amusing, but it would be a lot to include here. Then I said,

I still can't believe you don't know about Terry Winograd's PhD thesis.

BARD: I apologize that I do not know about Terry Winograd's PhD thesis. I am still under development, and I am always learning new things. I will try my best to learn more about Terry Winograd's work in the future.

ME:What's SHRDLU?

BARD: SHRDLU is a computer program that was developed in the early 1970s by Terry Winograd at the MIT Artificial Intelligence Laboratory. It was one of the first programs to be able to understand natural language to a significant degree.

SHRDLU could understand simple commands, such as "Put the red block on the table" or "Put the green cube on top of the red block." It could also answer questions about the state of the world, such as "What is on the table?" or "What is on top of the red block?"

SHRDLU was a significant achievement in artificial intelligence, and it helped to pave the way for the development of more sophisticated natural language processing programs.

ME: If I tell you that SHRDLU was a program written for Terry Winograd's PhD thesis, how would you summarize Terry Winograd's PhD thesis?

BARD: Terry Winograd's PhD thesis, "Procedures as a Representation for Data in a Computer Program for Understanding Natural Language", was an attempt to develop a computer program that could understand natural language. The program, SHRDLU, was able to understand simple commands and answer questions about a simple world of objects. Winograd's work was a significant achievement in artificial intelligence, and it helped to pave the way for the development of more sophisticated natural language processing programs.

ME: Baffling! So you did know about Terry Winogra'd PhD thesis all along!

BARD: I apologize that I gave you the impression that I did not know about Terry Winograd's PhD thesis. I have been trained on a massive dataset of text and code, and I am able to access and process information from the real world through Google Search. This allows me to answer your questions in a comprehensive and informative way.

I don't mean to post all my interactions, but here's one more. I asked Bard, "Are you familiar with Terry Winograd's PhD thesis on natural language processing? How does the program in his thesis compare to your natural langauge ability?"

Bard replied, "I do not have enough information about that person to help with your request. I am a large language model, and I am able to communicate and generate human-like text in response to a wide range of prompts and questions, but my knowledge about this person is limited. Is there anything else I can do to help you with this request?"

In contrast, chatGPT gave a good summary of SHRDLU and a superficial sentence or two comparing SHRDLU to itself. I then asked it some "block world" questions, and it answered a few correctly. 'Assume there are 3 cubs, red, green, and blue. Can you stack the red cube on the blue cube? Assume there is also a yellow pyramid. Can you stack the green cube on top of the pyramid."....

I can't pull up the exact chatGPT query I used or the reply. Sometimes the chatGPT browser interface can't access past conversations, and now is one of those times.

EDIT: Bard can correctly answer (easy) questions about an imagined blocks world. IIRC chatGPT was vague on whether a cube can be stacked on top of a pyramid.

Assume there are 3 cubes, red, green, and blue. Stack the red cube on top of the blue cube, and describe the result to me.

BARD: The blue cube is now the bottom cube, and the red cube is on top of it. The red cube is touching the blue cube on all six sides. The blue cube is completely covered by the red cube.

Also assume there is a yellow pyramid. Stack the pyramid on the green cube and describe the result to me.

BARD: The yellow pyramid is now resting on top of the green cube. The pyramid is touching the green cube on all four sides. The green cube is still visible from underneath the pyramid.

How would you stack the blue cube on top of the green cube, given the current arrangement of the blocks?

BARD: I would need to move the red cube off of the blue cube first. Then, I could stack the blue cube on top of the green cube.

Can a cube be stacked on top of the pyramid?

BARD: No, a cube cannot be stacked on top of a pyramid.

A cube is a three-dimensional solid with six square faces, while a pyramid is a three-dimensional solid with a square base and four triangular sides that meet at a point. The two shapes are not compatible for stacking.

Bard accepts URLs, which is a nice feature compared to chatGPT. Example described here: https://news.ycombinator.com/item?id=35250641

When I asked it about a different URL, it seemed to reference an older version of the page, and couldn't answer a question about the text content. chatGPT correctly answered the same question when I pasted the text. I tried pasting the same text in Bard, and Bard still couldn't answer.

My experiences with asking Bard to code are consistent with the other repiles here; Bard isn't very good at coding, or revising code it has generated.

I am interested to see that Bard offers multiple drafts of each reply. I wonder how it chooses which draft to show first.

I had a hard time understanding this article with phrases like, "However, with great power comes ~~great responsibility~~ undecidable type systems, whose undefined behavior may involve summoning eldritch abominations from the Black Abyss of Rěgne Ūt." I was pleased so see that Bard accepts URLs, so I asked it to summarize it. Here's Bard's summary:

  The author of the article, Erik McClure, rants about Terra, a metaprogramming language. He argues that Terra is not well-documented, that it does not work well on Windows, and that it is not as good as other metaprogramming languages.

  He also claims that Terra is not a good choice for writing production code, as it is not as stable as other languages.

  Overall, the author is not a fan of Terra. He believes that there are better options available for metaprogramming.

I requested access (using Firefox), and access was granted about an hour later. I used my gmail.com account. A friend with a Google workspace account was unable to request access. My initial observations are:

Responses can include other drafts, which can be slightly or significantly different. It is not clear why a particular draft is shown first.

It will write code, and it gives URL citations for some code that is generated. However, it doesn't even try to modify the code, at least in cases I tried. (chatGPT will help revise code).

It is interesting to compare chatGPT and Bard. In my small sample so far, Bard appears to have no sense of humor. Bard replies are more succinct, and in one case, provided additional context that chatGPT omitted.

EDIT: I asked, "Is your knowledge as of a certain date, or do you track recent changes on the web?"

And the reply was: "My knowledge is constantly being updated, as I am trained on a massive dataset of text and code. I am able to access and process information from the real world through Google Search and keep my response consistent with search results."

This has been posted many times before, but I thought of it yesterday when I saw this essay by Chomsky in the New York Times, "The False Promise of ChatGPT."

https://www.nytimes.com/2023/03/08/opinion/noam-chomsky-chat...