Closi

> My own answer is that while some human speech behavior is possibly generated by systems that function in a semantically equivalent way to current LLMs, human cognition is capable of tasks that LLMs cannot perform de novo even if they can give the illusion of doing so (primarily causal chain reasoning). Consequently, LLMs are not in any real sense equivalent to a human being, and using them as such is a mistake.

In the workplace, humans are ultimately a tool to achieve a goal. LLM's don't have to be equivalent to humans to replace a human - they just have to be able to achieve the goal that the human has. 'Human' cognition likely isn't required for a huge amount of the work humans do. Heck, AI probably isn't required to automate a lot of the work that humans do, but it will accelerate how much can be automated and reduce the cost of automation.

So it depends what we mean as 'use them as a human being' - we are using human beings to do tasks, be it solving a billing dispute for a customer, processing a customers insurance claim, or reading through legal discovery. These aren't intrinsically 'human' tasks.

So 2 - yes, I do believe that they operate in a way that makes them useful for tasks. LLM's just respond to text prompts, but those text prompts can do useful things that humans are currently doing.

I think the most obvious drivers were:

* Time - Waymo started work on this in 2009 and are now getting a faster adoption in 2024. Uber started in 2015 and sold it off by 2020. Time has given Waymo the advantage of slowly ramping up and starting with toe-dips.

* Funding - Waymo has spent more money to get where they are now (probably 2-3x what Uber spent in total). Google also has deeper pockets than Uber, which means that there is less pressure on quickly ramping up Go To Market or immediately getting profitability.

* Culture - Waymo was much more cautious (likely because of funding structure) which is now paying dividends in terms of regulatory approval and consumer trust ('sometimes you have to go slow to go fast').

I think simplifying her to 'youtube influencer' is unfair - she is a doctor of theoretical physics with a specialism in quantum gravity who produces science content for youtube. She knows the field enough to comment.

She doesn't even say that this isn't a big leap (she says it's very impressive - just not the sort of leap that means that there are now practical applications for quantum computers, and that a pinch of salt is required on the claim of comparisons to a conventional computer due to the 2019 paper with a similar benchmark).

Historically there have always been ads in paid news media (look at the newspapers, they have had ads in print since almost their inception) so I don’t think there is anything out of the ordinary here…

I.e. for paid print media the ‘correct’ amount of ads is almost definitely non-zero from the publishers perspective.

(Personally I’m not too bothered about ads in a paid online website as long as they aren’t excessive or intrusive but people may disagree and can vote with their wallets)

In my experience you are right - IT will always deny these requests, so you need to build the solutions in a way that avoids accessing existing DBs.

Usually it’s replacing a spreadsheet, so either the information can be manually keyed in or can be imported from various reports. Sometimes you even get into screen scraping, sometimes scheduled reports that are getting dumped to a drive and getting imported… basically any way that avoids needing to get permission from the IT team.

100% this! In some companies the 'simple app' that is described in this post will get some ridiculous quote from central IT/tech ('it will take our team 4 sprints') and then never get signed off. IT will also ban anyone spinning up their own servers due to support issues.

No code platforms manage to get around this.

Another use case - I work for a 'non-tech' consultancy. Clients typically won't like paying us to spin up some flask/django/rails app, but are happy to pay us to spin up some sort of no-code thing for them (perception is that it will be easier to self-support, which is also probably the reality compared to me developing some sort of rails app and then leaving the company).

I don't have firm thoughts on the price point, but two examples of real-world use cases would be:

* There is a food production company where their QA's do a monthly walkaround. It takes approx. 2-3 hours to type up notes after the walkaround. I'm in the UK and QA's are paid approx. £32k GBP, so 3 hours of their time is more like £50 benefit.

* Lots of logistics companies do daily walks with shift/team leaders. While these aren't usually typed up or anything, it would be great to document them in terms of actions and a tasklist to complete. The alternative to the software would be getting a team leader to write up notes after the walk, and this would take maybe 30 minutes. A team leader might be £28k p.a. so cheaper to get them to do it than buy software at $12k p.a.

The cost of the software would need to be a fraction (e.g. 10%) of what it is at the moment though for these sorts of use-cases to pay off.

Maybe a more generic version of the software not targeted at the construction niche could be something like £49 per month per user? Sounds more like the sort of level I would expect.

But I'm thinking that $1k is like way way way out of the reasonable range of my use case, and this is so different to your current business model I imagine it's irreconcilable.

FYI - this could be really useful in logistics operations and production too! (Which is my background, although I suspect the price point is unfortunately much too high for that application).

> Can you think of anything?

Yes, lots! £100m could give a hospital a substantial renovation for example.

> Firstly, the 'bat shed' (officially SWBMS) is expected to cost £100m. This is neither expensive nor wasteful for a structure nearly 1 kilometre long and "designed to accommodate up to 36 high-speed trains passing through the structure every hour of operation for 120 years, plus frequent conventional rail traffic in addition" as reported by Architects' Journal[1].

It might be good value for a 1km tunnel (or not, I don't know) but I think this argument misses the wood for the trees.

The main point is more "should we be spending £100m on a bat tunnel?"

i.e. What else could £100m of public money buy us, and would it be better than a 1km bat shed?

I suspect over-engineering and being allowed to generally spunk money up the wall was the main culprit.

The £100m bat shed isn't a sign to me of over-zealous environmentalists, it's a sign that the project was mismanaged because there wasn't enough pushback on spunking £100m up the wall with a mindset of "oh well, it's a big project, I guess £100m isn't much in the scheme of a project in the tens-of-billions things!.

Agree with this. A few prompt variants:

* What if you allow the model to do Chain of Thought (explicitly disallowed in this experiment)

* What if you explain the board position at each step to the model in the prompt, so it doesn't have to calculate/estimate it internally.

You are right - it's a balance and definitely a matter of taste.

Although not-ornate doesn't necessarily mean dead-simple or bad. For instance compare the following:

> The reverberations of the trumpet stirred the knights from their deep repose, igniting a tumultuous awakening.

With a very similar sentence from Tolkein:

> At that moment, among the trees nearby, a horn rang out. It rent the night like fire on a hill-top. Awake! Fear! Fire! Foes! Awake!

This is much less ornate, with simpler language, yet easier to parse and the image is much more vivid.

Pretty amazing effort, this looks like a great labour of love!

I can't give feedback on the code/technology, but on the writing on the lore section, I would try to simplify the writing. For instance the following:

> The reverberations of the trumpet stirred the knights from their deep repose, igniting a tumultuous awakening. With swords unsheathed and hearts ablaze, they clashed in a thunderous symphony of war, each seeking to claim dominance over the waking realm.

Feels too ornate (purple prose) and could be more directly put as:

> The trumpet’s call jolted the knights from their rest. Swords drawn and hearts alight, they clashed in a fierce battle, each striving for dominance.

I'm not an author or anything, but a little bit of copy writing could help - although this might just be me as it's probably a matter of personal taste!

It does! Drag some controls onto the form and hit the play button.

I think this is too loose of a definition of code...

Yaml isn't really code in the same sense as python is, and Shakespeare didn't write in code in the same sense that the diagrams in this article are.

Why not write recipes in code too?

Or a screenplay?

(Because there are better formats for them)

I don't think the 'craze' is thinking LLM-based AI will be the singular technology that changes everything.

The craze is that all combined breakthroughs across all types of AI/ML, including techniques that have not yet been imagined, represent a theoretical near-future technology that changes everything.

Besides, 10-15 years is nothing. I don't think 3D printers are a truly transformative technology compared to AI, however let's remember that WW2 aside, it took both airplanes and computers about 30-40 years until they had a broad societal/consumer impact (excluding military uses)

They could say that, but they haven’t to me in the past.

Of course people can say no to things, but that doesn’t mean that you shouldn’t ask when there is a good probability that they would say yes (particularly if you ask for the right amount in the right way) and very little downside.

Maybe instead of:

> [Me]: I’d like $Y

You could try:

> [Me]: That matches my base salary at the moment but my current employer gives me healthcare and a car allowance so I would be out of pocket if I moved from my current job. In addition I currently am eligible for a 10% bonus. Would you consider an additional $5k/yr to cover some of this? I would love to move… although I can’t justify a pay cut, but think this is a really good opportunity and would love to work with you so hope we can make something work.

Ie you gotta give them a reason to justify paying you more, and to think that paying more is required for you taking the job. Also you have to be realistic in how much you ask (smaller amounts are more likely to be accepted, larger amounts pay more but are more likely to be rejected outright as if something is too large you can’t really counteroffer)

I suspect this isn't about making something look 10% better - it's about making something look 0.1% better in some circumstances.

Which is totally great! The world needs lots of 0.1% improvements because 100's of them can add up to make something look or feel better when applied at the right time in the right way.

It's because current AI is meh on a phone.

Once someone manages to combine Siri's phone integration + ChatGPT Voice mode and it runs locally I will be very interested. Until then, I'm a bit meh on it.

Next step they should put a minimum price on non-disposable vapes at £50-£100.

Also cost per ticket gets more expensive in the higher price brackets, which you wouldn't expect!

FYI - Your website doesn't work on Firefox (the top nav bar) :)

CRM sounds interesting.

ChatGPT-O1 is quite a bit better at certain complex tasks IMO (e.g. writing a larger bit of code against a non-trivial spec and getting it right).

Although there are also some tasks that Claude are better at too.

It's not a class thing.

Fox hunting, horse riding, polo and skiing... yes. Conkers, No.

"Now we have dismissed Tesla, we have now decided to purchase tiny clown cars, as we could squeeze many police and suspects in each car"

2 months later

"We have now found that clown cars actually don't go fast enough, as they are only designed to travel at 2-3 miles per hour in a circus ring. Also they are not as big internally as you might think. We are now investing in a drag racing car that will be retro-fit for police work."

Totally agree - this story is only really interesting because it is Tesla, and it seems like they could have worked out that they wouldn't be suitable ahead of time. But equally this looks like a trial and maybe they just had to do the trial to dismiss it as a silly option.

It would be the same if some police force decided to purchase BMW Mini's, spent $100k retrofitting each for police work, and then complained that they were too small after the fact.

Total own goal!

It's not really an issue with EV's - if this was a requirement I'm sure they could have modified parts of the car to provide protection.

The issue seems to be that the purchase was poorly thought out and didn't think about all the requirements of a police car.