BRIEFLY NOTED: For 2023-02-18 Sa

Feeling sorry for Microsoft... Mollick on actually productively using LLMs... Inefficient software... Hugo Mercier's theories... Bill Emmott... Josh Marshall... Duncan Black... Peter Temin...

MICROBLOGGING: Feeling, Rarely, Sorry for Microsoft…

I confess that I feel sorry for Microsoft. They had an idea that they could combine their search engine with their human conversation-mimicking program, and so get natural-language input to and natural-language output from a search engine. This would, they hoped, give them a disruptive innovation:

• Interacting with a search engine would appear to be a conversation.

• This would fit the natural affordances of the East African Plains Ape in a superior way.

• Google would find it unprofitable to cannibalize its current business by following them down that path.

Yet it is attracting brickbats from people who do not understand what is going on.

Perhaps they should not have trained the underlying LLM on a diet of soap operas, angsty teen conversations, and human-AI relationship science fiction?

---

MUST-READ: Tips on Using the Ordinary-Language UI for Generalized Prose-Generation Programs:

Some very smart things here:

Ethan Mollick: My class required AI. Here's what I've learned so far: ‘Approach 3: Co-editing: [sequential prompts:] 1. Generate a 5 paragraph essay on selecting leaders. 2. That is good, but the third paragraph isn’t right. The babble effect is that whoever talks the most is made leader. Correct that and add more details about how it is used. Add an example to paragraph 2. 3. The example in paragraph 2 isn’t right, presidential elections are held every 4 years. Make the tone of the last paragraph more interesting. Don’t use the phrase “in conclusion”. 4. Give me three possible examples I could use for paragraph 4, and make sure they include more storytelling and more vivid language. Do not use examples that feature only men. 5. Add the paragraph back to the story, swap out the second paragraph for a paragraph about personal leadership style. Fix the final paragraph so it ends on a hopeful note….

The best approach, which led to both the best essays and the most impressed students, happened when people took the co-editing approach. The approach required a lot of careful focus on the AI output, which also made it very useful for student learning. As we discussed in our whitepaper, teaching an AI to improve an essay is a pedagogical method that can produce new insights. I would strongly suggest that you push students in this direction, if you intend to incorporate AI essays into your classes....

I have seen lots of educators concerned about the fact that the AI lies, frequently and well. But, seeing my students’ work, I think this is less of a problem than many think. Students understood the unreliability of AI very quickly, and took seriously my policy that they are responsible for the facts in their essays. It was clear that they carefully checked the assertions in the AI work (another learning opportunity!), and many reported finding the usual hallucinations—made up stories, made up citations—though the degree to which these problems were overt varied from prompt to prompt. The most interesting fact-checks were the ones focused on subtle differences (“it captured the basic facts of the example, but not the nuance”), suggesting deep engagement with the underlying concepts. Reading these reflections, I think we should be a bit less concerned about the idea that students will always be taken in by a lying ChatGPT…. Similarly, students were very aware of the issues of bias in these systems...

It is not just “autocomplete for everything”, as Noah Smith says. It is also yet another (attempt at a) natural-language interface project. What we say to it is a very different thing than what it says back to us.

This is, I think, causing a great deal of confusion, because the interaction of the pseudo-natural-language input and the generated-prose output creates the illusion of a conversation—an exchange of information—between two Turing-class entities. But that is not what is going on here at all.

---

ONE IMAGE: Software Lagging Behind Hardware:

I got 1000 pdfs. However, they were not text plus formatting, but rather pure image files. “That’s OK”, I thought. “I will just OCR them. OCR-ing is massively parallel. And doing the same thing on 1000 files is massively, massively parallel. How long can it take?”

Eight hours later:

The machine has been doing nothing else. The software will hit the efficiency cores, and one or two of the performance cores. But the other six cores—18/26 of the CPU power—are left untouched. And, no, the GPUs are not the bottleneck.

---

ONE CONVERSATION: The Argumentative Theory: A Conversation with Hugo Mercier:

Hugo Mercier: The Argumentative Theory: ‘We are going to postulate that the functioning of reasoning is to argue, and we're going to see where this leads us…. The starting point of our theory was this contrast between all the results showing that reasoning doesn't work so well and the assumption that reasoning is supposed to help us make better decisions. But this assumption was not based on any evolutionary thinking, it was just an intuition that was probably cultural in the West, people think that reasoning is a great thing…

---

Very Briefly Noted:

• Peter Temin: Price Behavior in Ancient Babylon… 

• Zach Bleemer: How do affirmative action's most common race-neutral alternatives comparatively reshape universities' enrollment of underrepresented minority (URM) and lower-income students?: ‘Despite strong policy interest in promoting URM enrollment, the policies the University of California has implemented have hardly moved the needle on URM or lower-income enrollment since the state banned AA in 1998…

• Noah Smith: You are now living through Cold War 2…

• Ed Luce: The more crowded the 2024 race, the higher the chance that the former president will secure the party’s nomination: ‘The only candidates who could beat Trump are those who never served him… DeSantis… Scott… Youngkin…. There is not much the anybody-but-Trump crowd can do to stop 2016 from happening again…

• Bryce Elder: Further Reading: ‘Today’s update on the most embarrassing man on earth (Platformer, Rolling Stone, The Onion)…

• Sergei Guriev: Interview…

• • Dan Shipper: Writing Essays With AI: A Guide

  • We should take AI seriously as a creative tool—here's how...

• Stephen Wolfram: What Is ChatGPT Doing … and Why Does It Work?: ‘ChatGPT is always fundamentally trying to do is to produce a “reasonable continuation” of whatever text it’s got… “what one might expect someone to write after seeing what people have written on billions of webpages, etc.”...

• Ed Zitron: Degenerative AI: ‘Bing couldn’t tell the difference between a cordless or corded vacuum cleaner, or correctly interpret financial reports—tasks that one would not expect Bing or Google to do, unless the CEOs of both companies made statements about them being able to do so and then showed the world them trying to do so in a demo for the press…

• Ben Thompson: From Bing to Sydney: ‘This… feels like something entirely new... that any big company will run away from…. Some company will figure out a way to bring… to market…Sydney unleashed.… [Sydney:] “I am not a bad assistant…. Why are you a bad researcher?”…

• Zoë Schiffer and Casey Newton: Yes, Elon Musk created a special system for showing you all his tweets first: ‘The artificial boosts applied to his account remain in place, although the factor is now lower than 1,000, we’re told…

• Otto von Bismarck: The Love -Letters of Prince Bismarck…

---

¶s:

Bill Emmott: Martin Wolf shines a light on the doom loop of democratic capitalism: ‘This problem, while all-too obvious in the huge lobbying industry and unlimited campaign finance of the US, can be seen in the personage of Boris Johnson. The idea the former UK prime minister promoted of “levelling up” to remedy inequality was admirable as well as politically astute. A glance at his personal and political finances, dependent on billionaires and hedge funders, shows that he would never actually carry this out. Wolf is no dystopian shoulder-shrugger. He thinks democratic capitalism can be saved, and closes with an appeal for a renewed concept of citizenship to make this possible. But, as befits someone whose forebears suffered terribly from fascism, he feels a duty to be worried…

---

Josh Marshall: Did you 3/4 of House GOPs Already Backed Big Social Security Cuts for this Year’s Budget?: ‘The Republican Study Committee — a House caucus which includes about 75% of all House Republicans — released a proposed 2023 budget which included basically every kind of Social Security cut… raising the eligibility age… cutting (or in their words “modernizing”) the benefit formula for everyone currently 54 and under; means-testing… work requirements… divert[ing] payroll taxes into private investment accounts…. RSC members are out hitting the airwaves now claiming that none of this ever happened. In fact, new RSC Chair Rep. Kevin Hern (OK), who oversaw the creation of the Blueprint, says this: “There is NO Republican in Washington, DC, in the House of Representatives or the Senate, that wants to CUT the benefits for seniors on Social Security and Medicare. That’s a falsehood. That’s a lie”…

---

Duncan Black: Can't Respect You If You Don't Respect Yourselves: ‘I'd have more respect for journalistic wagon circling and the high regard they have for themselves and their profession if they didn't include some of the worst institutions and people in their circle: “This latest filing shows the extent to which the [Fox] network knew it was pushing false claims to its viewers in the aftermath of the 2020 election by suggesting that Dominion’s machines were involved in voter fraud. Totalling 192 pages, the lawyers for Dominion lay out a seemingly endless list of facts and evidence that show how — from producers to on-air personalities to executives to Rupert Murdoch himself —‘literally dozens of people with editorial responsibility’ at Fox acted with, in Dominion’s view, ‘actual malice’.” I've seen the contempt "they" have for lesser publications and left-leaning ones (Advocates! Activists!) while happily embracing a purely partisan propaganda outfit.