Closi

Yes, also it says that the account management team had been in contact within the week, just that they didn't have a 'legal/corporate' response within the week, which is also not an unreasonable length of time as those things can take a bit of time.

> So if they no longer have a contract, how is it legal for Flock to continue to use the allocated resources? Why not hit them with theft of service for continuing to draw power and use public infrastructure?

Well we don't know what was in this contract was that they had with Flock in the first place, but usually a contract termination like this would have some requirement to give a reasonable notice to allow Flock to disconnect cameras etc, and a week probably isn't reasonable notice.

I think this is sensationalised - it looks like they only waited about a week after cancellation before they started unplugging them.

But I can't find any easy information on the contract they had with Flock, but it could well take a week for Flock to be able to respond properly after working out what to do with lawyers etc. The article says that the account management team had been in touch... so... I think this is just a PR piece for the police chief to look 'hard'.

Its easy to see what works in hindsight, but in reality with this sort of research you have to find an approach that’s working and keep pushing it and see where it goes.

The results so far from genie have been stunning, and its not obvious to me that another approach is guaranteed to end up better for all use cases.

I mean Astra actually means star, but putting that aside, we also have GPT sunburst and GPT flare - how does the logic apply to these?

Is a sunburst supposed to be bigger than a flare? I mean you get solar flares too? Or is a solar flare bigger than a sun 'burst' which usually just means some sun coming through clouds.

100%! The other problem is that the next model release from OpenAI there will be another different 4 terms.

It depends if the goal is a more accurate simulation, or if the goal is a tool for creatives to build games or content.

If it's the latter, I don't see a 'hole'.

> In a few years our names would have to advance to labels like giant, behemoth, gargantua

You wouldn't - you would just put the version before it - 6.1 Small, 6.0 Medium etc. Not that complex.

Think of an iPhone 18 Pro - No need for 'iPhone 18 Even Better Super Pro' because the 'Pro' is relative with the version/generation.

At least there is a simple system - the words 'Mini' and 'Max' are intended to refer to the size of the item (that one is miniature, and the other is their maximum size).

If they were called iPhone 14 Luna and iPhone 16 Sol then this is less clear.

Or on the flipside, if the models were called GPT 6.1 Mini and GPT 6.1 Max, nobody would be arguing about the names being hard to grok. The words 'mini' or 'small' are more descriptive than 'Luna' or 'Terra'.

> I work at OpenAI and I can try to explain our logic [...] general, models that are bigger will be more performant.

> If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.

How about naming the models:

"Small", "Medium", "Large" and "Very Large"

Or refer to them by parameter count like other companies (e.g. this is 6.1 5T)

If I have a clothing store, and rather than name the clothes sizes Small, Medium and Large I decide to call them Urgon, Planfar and Fangor I shouldn't be surprised if people think my size naming is confusing or silly. And IMO it's not a great explanation to say "oh it's easy, if you are skinny then you should get Urgon, but then if you are a bit more of a medium build you should get a Planfar, but if you want something a bit baggier than Fangor will really work out for you"

> hopefully this quells some of the "AI is unprofitable everyone is spending a dollar to make fifty cents" arguments.

I think everyone has known that inference is profitable overall for the major providers for a while (I'm sure not all user accounts, but in aggregate).

The major providers are unprofitable after you add in the training and research costs and other overheads.

This is why the open source models are such a threat if they become good enough - it will destroy the business model of just providing raw chat/intelligence.

Arguably that's also why a few of these providers are trying to diversify into whole-platform approaches rather than just relying on having their whole thing be a great model.

IP Whitelist, this is standard stuff.

I think this is quite a theoretical view, and in reality:

* The majority of missed payments and defaults don't end up as CCJ's, as often these are settled or just happen without going to court.

* Fraud being illegal doesn't stop people committing fraud on you. The idea is to prevent yourself from getting frauded.

Without some sort of central authority, surely this is just vulnerable to a rollback attack?

i.e. if I am a bad actor, and had just defaulted on a loan, why would the next document I send not be my copy of the previous 'clean' record before I got that loan?

Or what prevents me from sending out the same copy of the document to 10 other parties and getting 10 loans concurrently?

These are the sort of scenarios that the existing credit agencies prevent.

Then we follow it through and get to some sort of decentralised ledger to protect against these attacks, but that's then vulnerable to not being able to be expunged (how do we make sure it's possible to correct the record, as the CRA's are legally required to do when information has been written incorrectly).

So we end up with some sort of central authority.

Well, it's usually much cheaper to use a non-Amazon 3PL to ship your Amazon volume (along with other D2C channels), but then you might not have the prime badge if you go on the cheaper side of the market.

And even if you want a prime badge, it's often/usually cheaper to go with a 3PL that can do some sort of SFP setup if you have reasonable volumes.

Although Amazon has a stranglehold on 3PL's too (or if you are a seller offering SFP), as they will force you to use Amazon Logistics at a certain % of volume if you want the prime badge which can be more expensive than other carriers.

(I work in logistics consulting and do network strategy design & 3PL selection as part of that so have a reasonably up to date view on comparative/competitive costs for this sort of thing - mostly in Europe though and appreciate the USA market is different)

I don't see how this works in reality - The issue is that users wouldn't upload negative information (e.g. county court judgements or the bank robbery from your example).

Credit isn't like insurance where these things can be checked later during a claim (which can then be denied) - you need to know the person is a bad-payer before you send them the ten thousand dollar loan as that affects the terms, not afterwards when they default and you can't get your money back.

It's much faster and cheaper (an order of magnitude).

And theoretically will give you better answers statistically as it's calibrated.

Or almost any invention for that matter.

Washing machines can’t do anything humans do, they just remove labour. Trucks don’t do anything longboats can’t, it just need less labour and time/effort to build roads rather than canals. Computers can’t calculate anything humans can’t dry run by hand etc.

Everything in reality is about reducing time/effort/material/cost or achieving more with less resource.

And it's not like it was a load of detail either, it's a comment on a blog 16 years ago.

You don't get points for throwing out a possibility for the type of cipher, you get points for solving it.

Yes! And all the splash screens!

Or even just the computer taking 5 minutes to start up.

> Sorry, you are completely off. My machine 20 years ago was not an SOC and didn't have an SSD either. It was fast as lightning to open a window.

On my machine 20 years ago most apps had some sort of loading splash screen, so I think you just have rose tinted glasses imo!

You don’t really get those anymore - you click and the app is open.

Nice idea and appreciate it's a prototype - appears very glitchy for me in Chrome where bus routes appear/disappear when zooming in/out, i.e. it can show no routes at all, but then if you either zoom in or out they appear (some - not all).

Not sure if just me / my browser but thought I would share.

This is in Edinburgh btw - not sure if it's just an issue with that city.

> In principle you don't need 32-core GHz CPUs to move a window around on a screen without it feeling sluggish

I think the snappiness is less this, and more about the speed to open a file, how quickly chrome pops open when you open it etc. I assume this is a combination of hardware and how fast SSD->RAM is due to the SOC, but software will certinaly play a big part too.

> Not all values of "equally good" are equally good. Writing has rhythm, or at least it's supposed to, and synonym swapping compromises it.

We already know that a non-zero temperature improves quality though with current models (particularly with creative writing). The assumption that always picking the 'best' token results in the 'best' output is not the current reality.

And if you are already intentionally putting in randomness, I can imagine that it would be possible to seed the randomness in a way that is detectable but results in the same quality.

This is obviously not true for queries where temp = 0, but at temp = 0 then it becomes easier to identify anyway. I assume this technique implies some level of temperature.

I think the answer is 'it answers badly because of a lack of training data'

> Q: Whats the impact of asbestos?

> A: First, the job took 10 hours and 30 minutes. Next, let's change 30 minutes into part of an hour. Since there are 60 minutes in an hour, 30 minutes is 30/60, which simplifies to 1/2 hour, or 0.5 hours.

or

> Q: What is the five times table?

> A: We want to understand what the five times table is. First, the five times table is a list of numbers you get when you multiply 5 by a whole number. Next, the most common whole number you can multiply by to get 5 is 1. Then, the five times table shows the result of 1 multiplied by 5, which is 5. Now, let's check if 5 multiplied by 3 equals 5. Yes, 1 × 3 = 3, and 3 × 3 = 9, which is 5. Therefore, the five times table lists the result of multiplying 1 by 5. Answer: 5

Interesting project though! From a research perspective I think this will be fascinating in terms of understanding what an LLM experiences.

I've just tested it and it works - you have to disable touch though to get the full 'lock down to one event ticket'

100% - I would have assumed the author would have searched first as it's trivial to find this feature if you look on Google or ask ChatGPT.

I do wonder if part of the issue though is that we have thousands of these 'great' features, so discovery is poor because people can't possibly know about them all. 'great' features lost in a sea of other 'great' features.

Depends how much you are using it - for a casual user the $20 tier gives you just about enough use for hobby projects etc.

Or if you are doing full time development then you just need to pony up for the 20x tier if you want to use it.

Running Codex locally removes so much friction and is so fast compared to either running it on the web (where it has to build environments online etc which consumes time) or compared to copy-pasting chatgpt that it's just a no-brainer.

I mean it was just an example, but I get AI to do things like this all the time... ChatGPT planned my latest work trip by looking through flights, hotels and pulled together 3 options of itineraries for me, and wrote an email I could send to my client with estimated costs etc.

You can change the prompt, and give the agent tools, and that keeps the agent up to date.

The agent doesn't know the date, or know what hotels there are in Montreal, it sees:

> System: You are an AI agent. The date is 11th August 2026. Your knowledge cut-off is March 2023. User is based in <date>. If you need to search for something to support the user say {search:<term>} and a list of options will be provided along with instructions on how to access. Or say {help} for a full list of commands.

> User: Can you help me find hotels in Montreal for next weekend?

The AI then interacts with the tools given in the base prompt, which can obviously be updated. So it then goes:

> AI: Of course, let me search for that. {search: hotels in montreal for 16th August}

> System: [Provides list of websites]. Say {read[n]} to read option or say {start subagent:<goal>) to register subagent.

> AI: {start subagent: List hotels on booking.com available on 16th August}.

[etc etc, then eventually]

> AI: Yes I have searched for you and I found a few options!

While you can't change embedded knowledge, a good model knowing that the date is 5th January 2040 can infer certain things (e.g. while it might not have been trained on certain deaths, it can probably guess that it should search before answering if it means a person would be 102 and their last information is from 2024)