The Ai thread

[Huntn]

Many of them openly use it now, especially for imaging, and I can't blame them for that.

The creeping issue could be the loss of skills which is true of all automation.

 

[Eric]

The creeping issue could be the loss of skills which is true of all automation.

I think I mentioned this before but I recently saw a pulmonologist on the news who said it was his job to look over x rays to detect pneumonias, something that takes time and an experienced eye. He said after they set up AI that it was able to detect it within seconds with pinpoint accuracy, he also laughed and said he was out of a job. To me, this is where this technology can really benefit the medical community.

 

[Huntn]

I think I mentioned this before but I recently saw a pulmonologist on the news who said it was his job to look over x rays to detect pneumonias, something that takes time and an experienced eye. He said after they set up AI that it was able to detect it within seconds with pinpoint accuracy, he also laughed and said he was out of a job. To me, this is where this technology can really benefit the medical community.

See that is the challenge, how a for-profit system, takes care of its citizens with wholesale technology based job loss.

 

[Katbel]

Many of them openly use it now, especially for imaging, and I can't blame them for that.

If it’s for improving of course, not for cheating in the exams or for resolving school problems . You need to practice using your brain first.
I know it first hand: one of our boys, grade 6th has learned to find the solutions without using his brain , asking ChatGPT
Well it’s been brief. We are not kidding about that. Thorough explanation and in short : more homework with a lot of calculations with pencil and paper done in front of us. Teacher agrees and we too... but it’s like to try to stop a tsunami with a cardboard box.

 

[fischersd]

I'm very shocked that someone hasn't created a bank or insurance company "in a box". It's all math / analytics. Something that AI does very very well.

Whoever does it first will put the others out of business....and likely put millions out of work.

 

[Huntn]

I'm very shocked that someone hasn't created a bank or insurance company "in a box". It's all math / analytics. Something that AI does very very well.

Whoever does it first will put the others out of business....and likely put millions out of work.

You might say we deserve this as a painful lesson, not that the powers and Money would allow it.

 

[Nycturne]

AI has some good qualities, I have a photo manipulation program that uses AI and it’s pretty spiffy. This helps me a “know nothing” fix up old family photos pretty easily. Including enlarging photos. So it has a role.

I think it's worth pointing out there's a *huge* difference between the sort of ML models that enable faster edits, identification, etc... and LLMs. One of the things the LLM companies are trying to do is conflate LLMs with all AI so that someone who looks at LLMs and goes "WTF?" is seen as some sort of anti-ML luddite.

Image generators for example, are super-charged denoiser models, similar to the sort of stuff I use for astrophotography. The difference is tagging the sample data set such that it uses the tags as a "guide" on what is in the noisy image. And then you feed it pure noise along with the tags. It will then try to denoise (literally 'see' something that isn't there) an image from pure noise.

I'm very shocked that someone hasn't created a bank or insurance company "in a box". It's all math / analytics. Something that AI does very very well.

The catch there is that you also have a lot of logistics for when you still need to deal with cash, regulations, etc. And the fact that the math part absolutely must be deterministic. Something AI doesn't do very well. Humans aren't super great at being deterministic either, but that's why we discovered mathematics, to give ourselves the tools that our built-in cognition lacks. Excel didn't take over business because it made ledgers from vibes, but because it could be trusted to do the same calculation millions of times the exact same way, and could do it way faster than the human "computers" could.

You can use an LLM to maybe kick off certain types of analysis ("I see this pattern in the data"), but absolutely cannot trust the final analysis itself to the LLM, which means spending a fair bit of time building out everything you need. And when you've done that, you realize you still have something that looks like an existing bank, just with an LLM bolted on. Something existing banks are already trying to build. The moat is from all the stuff you can't let the LLM handle directly.

 

[Citysnaps]

There's a revolution coming. Governments seems incapable of expeditious action. Any government that attempts to thwart AI replacing jobs will only see those companies moving offshore to more hospitable bureaucracies. Capitalism has been inherently flawed....allowing a handful to attain untold wealth.

We've started to see the jobs being eliminated by AI. (IBM laid off about 8000 last year).

"So far this year, 49,135 layoffs have been attributed to AI", according to this article.

Capitalism (as it exists today) falls on its face when we no longer have enough people working to fund our governments and our economies. The US is already insolvent - owing far more than they can ever realistically repay....getting to the point that they won't be able to pay the interest on the debt.

The only way to keep this society spinning is to embrace Universal Basic Income and evolve our capitalism into something else. Reclaim the trillions held by a handful of people. Free education for everyone. Eradicate poverty and disease. Let's see how many Hawkings there are out there, when given the opportunity to learn.

We need a hybrid of capitalism that rewards people for their contributions to society. Maybe THAT should be the problem that we get AI to solve. In such a society, social workers and teachers would have a more lavish lifestyle than plastic surgeons that only perform liposuction and boob jobs.

Because we have seen all too well how deplorable the teeming masses can be....I suspect we're far more likely to end up in a "Mad Max" world than a Utopia where everyone gets to follow their dreams. Well, provided we don't nuke ourselves and cover the world in radioactive glass first (keep Donnie away from the football!!!)

Happy Monday!

And Oracle not long ago laid of 30,000 employees (a friend of mine being one of them).

 

[fischersd]

The catch there is that you also have a lot of logistics for when you still need to deal with cash, regulations, etc. And the fact that the math part absolutely must be deterministic. Something AI doesn't do very well. Humans aren't super great at being deterministic either, but that's why we discovered mathematics, to give ourselves the tools that our built-in cognition lacks. Excel didn't take over business because it made ledgers from vibes, but because it could be trusted to do the same calculation millions of times the exact same way, and could do it way faster than the human "computers" could.

You can use an LLM to maybe kick off certain types of analysis ("I see this pattern in the data"), but absolutely cannot trust the final analysis itself to the LLM, which means spending a fair bit of time building out everything you need. And when you've done that, you realize you still have something that looks like an existing bank, just with an LLM bolted on. Something existing banks are already trying to build. The moat is from all the stuff you can't let the LLM handle directly.

Any of the pieces that you need an actual human to human interaction (ex branches where people come in to talk to someone), you'll still need to have - as some people just won't use their phone or an app to get things done....but they'll all die off soon enough.
Contact centres, you farm out.

Here's an example - the job of estimators. They review the damage of a vehicle and put together the estimates for the repair costs. The AI models developed by a local insurance company have a 97% accuracy rate (and, unsurprisingly, they can tally up the parts lists much quicker than their human counterparts, resulting in claims being processed much quicker. You only need a handful of estimators kept to review the results to ensure there's no gaps in the processing.
AI does the job from the photos that are uploaded to the system. Eventually the customers will also have the capability of doing so, before the car even goes to the repair shop.

I'm not saying AI does EVERYTHING that an insurance company does - but you already have software that does most of the heavy lifting - you only need AI for the other 10% that humans are doing today. The remaining cost of face to face interactions will be pretty trivial compared to the thousands of jobs you eliminate in the back end.

And, yes, the banks and insurance companies have been working with AI for 15? years. I think a startup that didn't have a vested interest in retaining employees would be much more aggressive in implementations.

 

[Huntn]

And Oracle not long ago laid of 30,000 employees (a friend of mine being one of them).

The CEG…

 

[Cmaier]

Is AI Profitable Yet?

isaiprofitable.com

Red is expenditures. Green is revenue. AI is just a giant transfer of wealth to nVidia.

 

[Eric]

May 26th Google is fully transitioning to AI.

"AI Only Search" they announced AI-augmented and likely AI-First search instead. For example a Google rep replied to the TechCrunch article which went viral across social platforms with this statement: “We’re continuing to display blue links on the search results page in addition to AI responses. If someone chooses to ask a follow-up from an AI Overview, or selects the AI Mode button in the Search box, then that takes them to AI Mode. It doesn’t happen automatically – people have to choose to navigate to AI Mode.”

By the summer Google will have built-in capabilities for users to create their own search agents inside of search results. Think of these like Google mentions, but for specific data points. The examples Google specifically gives are stock prices and apartments for rent that match highly specific criteria. Google has added other stuff to SERPs over the years like images, products, and 3D content. Obviously, search agents might be more impactful than the others.

From Google's blog post:
"With information agents, you can stay updated on whatever matters most to you. Your agent will intelligently look across everything on the web, like blogs, news sites and social posts, plus our freshest data, such as real-time info on finance, shopping and sports, to monitor for changes related to your specific question.

It will send you an intelligent, synthesized update, with the ability to take action. So if you’re apartment hunting, you can brain dump all of the exact requirements you’re looking for, and your agent will continuously scan for you, notifying you when listings meet your needs. Or if you want to know the instant any of your favorite pro athletes announce a sneaker collab, your agent will let you know when a new drop lands so you don’t miss out. Information agents will launch first for Google AI Pro & Ultra subscribers this summer."

TechCrunch stated: "This shift means that “searching the web” will increasingly be performed by AI agents rather than humans. Instead, people will focus more on acting on the information those agents provide instead of manually clicking links."

And: "Links will become an afterthought with the coming changes to the Search results experience, which builds on Google’s earlier launches of AI search features, like its short summaries known as AI Overviews and its conversational search, AI Mode."

Essentially, Google is saying they are rebuilding the search experience for bots to do the searching on behalf of humans but only if the humans want them to. Of course, they are likely to push this more and more and could ultimately become AI-Only but this I/O that is not what was announced.

Google Search as you know it is over | TechCrunch

Google is transforming Search from a list of links into an AI-powered experience filled with conversational answers, autonomous agents, and interactive interfaces — a shift that could further reduce traffic to publishers across the web.

techcrunch.com

 

[fischersd]

The scary thing is, Google will also be able to tune the bias based on the person's profile (if they're not already).

TikTok is already doing this. Firstly, they identified everyone that was a "leftie" - ie - NOT a Trump supporter. Everything that we posted was flagged and removed "for violating community guidelines". (because, YES, we should all be worshipping the orange nazi).

Well, now it seems those of us flagged as leftists can't post hardly anything?! Example:

If Sundar Pichai is a Trump supporter, then they could very easily flag profiles (if you're signed in during your search) and put your results into "lanes" that they've decided. So you will only get filtered content.

 

[Cmaier]

May 26th Google is fully transitioning to AI.

"AI Only Search" they announced AI-augmented and likely AI-First search instead. For example a Google rep replied to the TechCrunch article which went viral across social platforms with this statement: “We’re continuing to display blue links on the search results page in addition to AI responses. If someone chooses to ask a follow-up from an AI Overview, or selects the AI Mode button in the Search box, then that takes them to AI Mode. It doesn’t happen automatically – people have to choose to navigate to AI Mode.”

@Eric, how else will I know the instant any of my favorite athletes does a sneaker collab?

 

[Eric]

@Eric, how else will I know the instant any of my favorite athletes does a sneaker collab?

Priorities.

 

[KingOfPain]

Here's one headline I never thought I'd read. But as a comment states, the use of a Tolkien quote is most likely to directly criticize companies that named themselves after words invented by Tolkien (like Palantir).

Citing Gandalf, Pope Leo says we must "disarm" AI

In an age of AI, Pope looks for "artisans of hope."

arstechnica.com

 

[Nycturne]

Here's an example - the job of estimators. They review the damage of a vehicle and put together the estimates for the repair costs. The AI models developed by a local insurance company have a 97% accuracy rate (and, unsurprisingly, they can tally up the parts lists much quicker than their human counterparts, resulting in claims being processed much quicker. You only need a handful of estimators kept to review the results to ensure there's no gaps in the processing.
AI does the job from the photos that are uploaded to the system. Eventually the customers will also have the capability of doing so, before the car even goes to the repair shop.

I'm not saying AI does EVERYTHING that an insurance company does - but you already have software that does most of the heavy lifting - you only need AI for the other 10% that humans are doing today. The remaining cost of face to face interactions will be pretty trivial compared to the thousands of jobs you eliminate in the back end.

Which is where I bring up ML vs LLMs. What you describe here is something that they’ve been building with ML models for a decade or more. But it’s also that these sort of automated processors tend to lead to more messy outcomes because everything is more opaque and the company goes “ML model says X, go away” when it does get it wrong more often than not. This has been stuff we discussed before LLMs sucked all the air out of the room, by the models incorporating biases from the datasets and entrenching the sort of biases we already had.

And, yes, the banks and insurance companies have been working with AI for 15? years. I think a startup that didn't have a vested interest in retaining employees would be much more aggressive in implementations.

I never said it’s the people for banks, I said it’s the logistics and regulations. Bypassing that creates risks for folks trying to beat the big guys. Which sure, people will take on because they don’t know any better, but we’ve built up the guardrails in the banking system for a reason. One of the more recent examples being 2008.

 

[Katbel]

Today I read an article on the NYTimes where they analyzed creativity with AI. (Free article)
"In one study, he and his team examined personal statements from more than 370,000 students, and found that after ChatGPT became available, their essays suddenly used diverse and colorful language, but lacked truly creative ideas."

In a separate study, the team found that human-written essays offered up to eight times more new ideas than those produced by A.I.

For the first time in human history, we have a technology that can generate words separately from the thoughts they represent. When a chatbot writes, it is predicting the next word that is most likely to make a “good” sentence or essay, based on the text it’s been trained on. It can identify sophisticated and creative word patterns independently of whether the underlying ideas represent something new.

When teenagers write their own essays, the work reflects their thoughts and personalities, their attempts to make meaning of their experiences. When we search for words, we are sifting through the same brain networks that form connections between ideas.

 

[Eric]

Here's the breakdown, if you want to use Google extensively it'll cost you as "AI-powered Search mode" is now the default.

Traditional Google web searches remain completely unlimited. However, if you are using Google’s conversational AI models (like Gemini) or AI-powered Search mode, usage is based on "compute-used" models rather than flat prompt counts, refreshing periodically. [1, 2]
1. Traditional Google Search

• Limits: Unlimited for standard web searches.
• Details: You can make as many normal web queries as you like on Google Search. [1, 2]

2. Conversational AI & Search AI Mode

• Free Accounts: You can make up to 25 conversational prompts per day using the standard dynamic AI view.
• Google AI Plus / Pro Subscribers: Daily allowances are significantly higher (up to 250+ prompts per day depending on compute complexity).
• Google AI Ultra Subscribers: Grants the highest available limits for intensive research, deep thinking, and multi-modal tasks.

 

[Eric]

A friend and I were shooting an event in San Francisco last weekend and we went to a Starbucks to make our edits before posting it to social. During the process he asked me about video and image stabilization, I told him I use the DJI RC4 gimbal. I looked at my Instagram about a minute later and had several posts with the RC4 in it, like it was painfully obvious every 2 or 3 posts I would see a post about it.

This is something I've searched before months ago but it clearly knew I was talking about in on this day. Looking at my iPhone the only thing that had the mic enabled was my own build app.