TechTakes

I know this shouldn't be surprising, but I still cannot believe people really bounce questions off LLMs like they're talking to a real person. https://ai.stackexchange.com/questions/47183/are-llms-unlikely-to-be-useful-to-generate-any-scientific-discovery

I have just read this paper: Ziwei Xu, Sanjay Jain, Mohan Kankanhalli, "Hallucination is Inevitable: An Innate Limitation of Large Language Models", submitted on 22 Jan 2024.

It says there is a ground truth ideal function that gives every possible true output/fact to any given input/question, and no matter how you train your model, there is always space for misapproximations coming from missing data to formulate, and the more complex the data, the larger the space for the model to hallucinate.

Then he immediately follows up with:

Then I started to discuss with o1. [ . . . ] It says yes.

Then I asked o1 [ . . . ], to which o1 says yes [ . . . ]. Then it says [ . . . ].

Then I asked o1 [ . . . ], to which it says yes too.

I'm not a teacher but I feel like my brain would explode if a student asked me to answer a question they arrived at after an LLM misled them on like 10 of their previous questions.

Recently read Brian Merchant's latest piece on the upcoming election, and I felt like making a quick-and-dirty prediction:

If Harris wins, I expect there will be some pretty harsh regulation against Silicon Valley. Putting aside everything but simple political pragmatism:

1. Elon Musk's election antics and Trump's support from tech billoinaires have shown SV holds a significant amount of power over politics - power which will almost certainly prove a constant thorn in Harris' side. As such, it'd be in her self interest to kneecap the Valley ASAP.

2. Public opinion of Silicon Valley has taken a pounding over the years for a variety of reasons, with the AI slop-nami just the latest and most serious grievance the public has against them. Any tech regulation a Harris presidency makes (especially against AI) is gonna enjoy significant public support from day one.

Fuck it, off-the-cuff prediction: I expect resistance to Silicon Valley is gonna turn violent during Trump's term.

I swear to god post the map with lots of diagonal arrows one more time

today's a fucking trip and a half

in a mall I saw some Honor laptop[0] ads with "AI" pitch sauce slathered all over it

later walking between art galleries I saw a roadsign ad for a property sales company promising "more effective strategies using AI" (more effective strategies for.. selling.. houses..? I guess..? (x up for doubt))

and now I get this shit in my mailbox from someone who absolutely purchased a dataset with this address in it:

From: Ai Everything GLOBAL <newsletter@event.aieverythingglobal.com>
Subject: Join the AI Elite—G42, TII, DIEZ, Dell & More!

[0] - I didn't even know they were in ZA, but nope, 2-floor hanging banners and ads allllll over elevator doors and shit

hackernews having a normal one (EDIT: Then things got worse)

HN Title: The evolution of nepotism in academia, 1088-1800

In the top comments:

dash2 55 minutes ago | next [–]

Here's an interesting extract:

We find evidence of nepotism for 5–6.6% of scholars’ sons in Protestant and for 29.4% in Catholic universities and academies. Catholic institutions relied more heavily on intra-family human capital transfers. We show that these differences partly explain the divergent path of Catholic and Protestant universities after the Reformation.

This relates to an important paper providing evidence that indeed Protestantism was associated with scientific progress: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4389708

red016 3 minutes ago | parent | next [–]

i’d be more interested in jewish numbers than protestant or catholic

Questions To Which the Answer is No, an ongoing series:

Generative AI - Can it actually help with the day-to-day technical work of a practising engineer?

The upcoming Autumn Panel Discussion hosted by the UCD Engineering Graduates Association (EGA) will take place on Tuesday, November 19, 2024, at ESB's Headquarters. This event will explore several pressing topics at the intersection of engineering and artificial intelligence. Attendees will gain insights into current trends in AI, particularly the latest advancements in generative AI tools and platforms. The discussion will feature case studies that highlight the successful integration of AI into real-world engineering processes, offering practical examples of its applications. Additionally, the panel will address the essential skills that engineers need to effectively incorporate AI into their workflows, providing guidance on upskilling for the future. The event will conclude with a Q&A session, allowing participants to engage directly with the panelists and delve deeper into the ethical considerations surrounding AI in engineering.

No, not software engineers, engineer engineers. Yikes!

Seeing a worrying trend of people wanting to cosplay as La Resistance, going dark, hiding from the cops. "Fun" fact, the Gestapo was extremely good at finding, torturing and killing people in the resistance. If people only know you via encrypted messaging, who is gonna raise a ruckus when you're sent to a camp?

Less election dreariness, Dr Michael Cook talking about the oasis minecraft 'game'.

Personal rant: in my ongoing search for a replacement for ProtonMail after they pivoted to AI had me almost sign up with Tuta because, hey, they looked good and were on my radar originally anyway, when I found out that they do not offer any IMAP/SMTP access at all.

I mean, I get it, their whole thing is privacy and, yes, storing mail locally on my machine kinda undermines the idea of strong and impenetrable E2E encryption, but I should at least have the choice like I do with Proton Bridge. Because without SMTP Tuta is completely unusable for git send-email. I mean, yes, technically I could copy-paste the output of format-patch into the web client but, first, I am lazy and don't wanna do that, and second, from my experience it rarely works anyway because the clients do some encoding crap so that git am doesn't eat it without cleanup.

Meh. I guess I have to keep looking.