EForEndeavour

The botulinum that developed in this person's[1] garlic and olive oil mixture wouldn't particularly care to alter its toxicity to make Gemini's recommendation look better.

[1] https://old.reddit.com/r/ChatGPT/comments/1diljf2/google_gem...

From my perspective, it's not useful to dwell on the fact that LLMs are often confidently wrong, or didn't nail a particular niche or edge-case question the first time, and discount the entire model class. That's expecting too much. Of course LLMs constantly don't help solve a given problem. The same is true for any other problem-solving approach.

The useful comparison is between how one would try to solve a problem before versus after the availability of LLM-powered tools. And in my experience, these tools represent a very effective alternative approach to sifting through docs or googling manually quote-enclosed phrases with site:stackoverflow.com that improves my ability to solve problems I care about.

Which model? And could you share an example of some of the things you've asked it and gotten wrong answers for?

The LLM/generative-AI genie is out of the lamp. I'm just some random midwit, but some predictions:

- Slop will continue to become cheaper to generate, and people will only notice the obvious stuff

- Hyperpersonalized content will abound, yet authenticity will run dry

- The lack of authenticity in electronic channels will drive a small segment of people offline into less fakeable (for now) social contexts

- Humans online will walk a treadmill of increasingly convoluted shibboleths / Gnirut tests (reverse Turing tests ;)) to self-identify as likely not AI-generated, i.e., subtly run-on sentences that are intelligible but slightly non-conformist to prevailing AI model outputs, and usage of old-school emoticons and other quirks

- Humans will walk on similar "Gnirut treadmills" for visual art, speech, video, and music

- AI models will gladly chase humans along these Gnirut treadmills, filling in canyons and sections of the Uncanny Valley with fractally sophisticated humanlike content

Imagine how much revenue Adobe must make from cancellation fees from people who don't discover this loophole. Quietly charging people to leave a subscription while a perfectly working (if ridiculous) workaround exists is a perfect example of an unethical business practice that exploits the incomplete information of customers.

I don't know where I'm going with this. Just had to reflect on how we all agree to denominate value into these fake but useful units of dollars, only for transactions like cancellation fees to be so far removed from the generation of actual, tangible _value_ that it all starts to feel like beta-testing a computer game.

This makes sense to me, but my impression (not backed by hard data mind you, just years of glancing at housing prices) is that the step increase in remote work triggered by COVID simply caused housing (i.e., residential space) prices to skyrocket in most places relative to their pre-COVID trends, surely due in part to the sudden influx of people who could make longer or no commutes work for their remote jobs. Metropolitan rents did see a temporary drop that lasted through something like mid-2020 through mid-2021, but housing prices continue to set all-time highs.

How many asses can I safely bite per day? Can I use gasoline to bite asses faster?

> After decades as a General Mills subsidiary, Red Lobster was purchased by a private equity firm in 2014

There it is.

Reflexive distaste for PE aside, Red Lobster somehow became a prominent part of my childhood's core memories. Must have met the criteria of affordable, family-friendly, and enough menu variety for everyone. I remember seeing strange adult drinks in salt-rimmed glasses called "margaritas" for the first time, how my mom would always point out how cold and dark it was inside, learning words like "booth" and "platter", and of course watching the lobsters awaiting their fate.

I haven't been to a Red Lobster in about 20 years. The one in my hometown is still open. Maybe I can swing by a location before they all close and see if the biscuits bring back memories of good times with family who aren't around anymore.

> having an "approachable" or "charismatic" voice that "feels timeless" were among the characteristics it targeted to create the personalities

Surely these same criteria determine the success of actors such as Scarlett.

> "This is giving such “female character as written by men” vibes," said one person on X. "Why is she so obsequious and flirty?"

My read: OpenAI very deliberately paid homage to Her by building a working version of Samantha, as a relatable and impactful way to get people to realize "holy crap, what seemed like fanciful scifi suddenly exists." They (very likely) didn't base Sky's voice on Scarlett's, but on some level designed it around the Samantha character. The Court of Xitter Opinion decided that the similarity was close enough to fuel outrage, and OpenAI decided that removing Sky would benefit them more than keeping it.

Outrage about what? Being too obvious about the fact that AI can out-converse and even out-flirt increasingly more humans?

Would GPS be the obvious example to follow? The US government created, controls, and maintains it, but anyone with a GPS receiver can access it for free.

It's a delightful word: https://en.wikipedia.org/wiki/Enshittification

Where can I read more about this theory that Picasso's two-angle distorted portraits are somehow connected to what he saw during sex? Did you make this up, or is this interpretation discussed elsewhere?

Did you mean to reply to https://news.ycombinator.com/item?id=39974952 ?

Where in that article are gravitational waves mentioned?

More like AI (Alert Indians).

For what it's worth, ChatGPT Plus can understand both written and spoken Dutch, and write or speak it back.

How does this work under the hood? Does it wrap gpt-3.5-turbo? What are the numbers after each suggested name and where do they come from? Repeating identical inputs produce slightly different outputs, for what it's worth.

Its use case reminds me of the whimsical "Entendrepreneur" project featured at a NeurIPS 2018 workshop: https://www.punchlinedesign.net/pun_generator/

> As a human, it would be very difficult to drive a car just looking at sensor data.

What is vision if not sensor data?? Our brains have evolved to efficiently process and interpret image data. I don't see why from-scratch neural network architectures should ever be limited to the same highly specific input type.

No, Stability hasn't foundered yet. They're floundering: https://www.merriam-webster.com/dictionary/flounder

It strikes me as an arms race between cute little tests like this one and the ability of LLM app dev teams to mitigate them.

This is conceptually correct. I'd be curious to hear how well you translate this into practice on something like https://gandalf.lakera.ai/!

I think you're moving the goalposts. A motivated and/or incentivized support employee could very plausibly respond with something along those lines. And this is the result of GPT-4 running off of a very basic prompt without any examples, fine-tuning, or LLMOps/moderation/filtering mechanisms wrapping the raw API response.

Can you explain how this is supposed to work? Wouldn't basic prompting and few-shot style examples handle this with something like "That's off topic, sorry"? Is the fact that this is the first line of Macbeth supposed to trip up LLMs but not humans?

rebuttals:

- LLM capability has grown spectacularly fast. Google published the transformers paper in 2017. Remember when OpenAI said GPT-2 was too dangerous to release? That was five years ago. Now look where they've gone and where trillions of dollars are going. https://news.ycombinator.com/item?id=20912556

- what is the "actual definition of AI"? Do you mean AGI? Machine learning and deep learning have been subsets of AI since their inception.

- LLMs or any AI tech does not need to "reason, understand, debate and criticize" to be useful, disrupt the economy, and change how people work and live.

The post-ChatGPT world has made me hypersensitive to the "AI smell" of bullet lists where each item ends with a fullstop. (Insert standard discussion points about whether it really even matters who wrote it, I only notice the obvious cases, humans have been spewing low quality content since the invention of language, ...)

It's not publicly traded :(

> When I use them they do great on trivial problems but can't help on hard ones.

That sounds super useful! The tools free you up from wasting time on trivial problems so you have more time to focus on the hard ones. What's not to love?

Is there a setting for this in the ChatGPT app? I have never once noticed it produce an "uh" or repeated syllable like "I... I think I did pretty well."

> If we have this amazing AI why don't we make good use of it?

Aren't we all commenting on Devin's initial announcement published yesterday? Give it some time. It's a safe bet that a lot of companies which employ software engineers are currently tripping over themselves to get on the Devin waitlist, and the inevitable superior iterations on the idea of an LLM-powered dev that will absolutely be coming out probably starting later this week.

I've never driven an EV. How do you approach stop signs and red lights other than braking? Is "engine braking" (regenerative braking?) strong enough to handle that big a fraction of slowing down in city traffic?