thebaldcoder.bsky.social
68 posts
4 followers
6 following
Discussion Master
comment in response to
post
Come back Mr Mollick you're sorely missed
comment in response to
post
Love your posts here but come on Ethan thats a bit daft: you have loads of fans here that arent going back to X and arent going to join LinkedIn. You're literally losing out on audience because of a small minority of idiots. That is social media sadly and the public space you're putting yourself in.
comment in response to
post
But isn't meaning based on our use of language? Certainly one interpretation of meaning is that we can describe it.
comment in response to
post
Brexit massively damaged growth. The Tories tried to cover that up by importing 2 million people, which worked to a degree leading to growth flat lining. Labour now want to limit immigration while at the same time not fixing the root cause: Brexit. Deep recession is coming...
comment in response to
post
Says the Conservative party, that to save the party's supposed economic reputation after Brexit decided to import 2 million people to try and cover up said economic disaster.
comment in response to
post
I'm not so sure of that: Google definitely seems to be on a steeper advancement curve coming from very far behind a year or two ago to being arguably neck and neck now. If this trend continues I see Google pulling further and further ahead in the coming years. That's just in LLMs as well.
comment in response to
post
It'd be interesting to know the compute put into both. I guess we will never know until it becomes useless information.
comment in response to
post
Starmer needs to grow a spine and do a Carney: we have a service deficit with the US along with the rest of Europe. I'm sure a few heavy digital services tax hikes will get the attention of the US. He should stop all this nonsense of shit covered chicken deals and focus on rebuilding our tech sector
comment in response to
post
I have no idea where the first column comes from. My first guess is that it's just Dump's dislike factor.
comment in response to
post
We got whacked with 10% tariffs for NOT having a trade deficit with the US. This doesn't sound like a win in any way. By Dumps logic we were actually being robbed by the US and now we've been slapped with tariffs for the US robbing us!
comment in response to
post
But we dont have a trade deficit with the US and yet we got whacked with a 10% tariff. Sounds like weve done worse than everyone else as at least they were getting something in the first place (by Dump's logic that is)
comment in response to
post
Its the second column not the first
comment in response to
post
How about the far larger radicalisation of elderly people via GBNews etc and far right news feeds that pour every reported act of violence and death into their phones on an hourly basis to scare the wits out of them that they live in some hyper violent dystopia Mrs Olukemi Badenoch?
comment in response to
post
Your reporting has been impeccable again. Keep up the great work - I need you to keep me a breast of matters! Thanks!!!
comment in response to
post
To me Dump has in fact been duped into believing the Russian machine of disinformation. I'm not sure there are a lot of his ideas they're following; more Putin's.
comment in response to
post
I think the constitution is gone. When you do everything through executive orders, go after the institutions of law, ignore election results, then there is no real power in the constitution.
comment in response to
post
This is some mad sh!t, half the world aint visiting the US for the next 4 years...
comment in response to
post
If I go around to my neighbours garden and torch their shed, I will be taken to court and at the very least, made to pay for a new shed. I suspect I would also be put in prison for arson.
Why shouldn't Russia have to pay for torching Ukraine? If they hadnt we wouldn't be making them pay.
comment in response to
post
You obviously can make this shit up...
comment in response to
post
Get well soon!
comment in response to
post
Absolutely. How do we think they're going to manipulate the votes to stay in power for as long as they like?
comment in response to
post
Oh no he's not a Nazi but he is a Neo Nazi. Difference? None really but one lives today and the other is dead.
comment in response to
post
Thanks for this Simon your reports are the very best
comment in response to
post
It just sees things differently and has to store far more information because of that.
comment in response to
post
But that's not happening otherwise the model would be correctly identifying 9.9 > 9.11. Im assuming just increasing the model size would fix the issue or breaking up the numbers into individual characters so that fewer vectors need to be stored. Either way its not really that the model is stupid...
comment in response to
post
No the model 'sees' tokens. That is all it is fed. Thats why it makes mistakes about r's in strawberry because it doesnt see the word strawberry as we see it. Probably its the same situation here. See Andrej Karpathy's recent video.
comment in response to
post
You're fantastic Simon. You got me through the first Trump disaster and now we meet again in this disastrous situation.
comment in response to
post
And after that sending back the Statue of Liberty and erecting a giant statue of Putin in its place.
comment in response to
post
That's probably because it sees tokens not characters. It sees that 11 as probably a single token rather than two digits i.e. its a sight issue rather than an intelligence issue. We use tokens for compute efficiency and it's just a matter of time before we fix these problems.
comment in response to
post
3) and the current policymakers will undoubtedly put the onus on the AGI labs. This is going to get bad very quickly once we unlock properly working agents.
comment in response to
post
So you're saying Europe is in fact leading the AI race (where Deepmind is based)
comment in response to
post
I do feel like text to video is tangential to the whole AGI thing. There doesn't seem to be a thing that humans do that is close to this. Dreams/imagination possibly? Art isnt really this.
comment in response to
post
We won't be able to align AI to us as it needs different resources: electricity vs food. When it starts to compete for space it will want power stations not farms... It will outcompete us for space.
comment in response to
post
Eventually maybe, quite possibly even but not in the 'transition period'
comment in response to
post
No capitalism will survive humanity wont. Capitalism requires intelligence not humanity (although we obviously started it). Once there is an A.I. economy that out competes our human economy for resources thats when the sh.. really hits the fan
comment in response to
post
The difficult part is doing this in realtime but we'll get there its just a scaling/optimization issue.
comment in response to
post
What are sensors doing? A: feeding data into the model. What are effectors doing? A: reading output data from a model. There's nothing particularly special about that - its a relatively trivial step. You can think of the human senses including the nervous system as just feeding data into the brain.
comment in response to
post
People need to stop worrying about whether AGI has been achieved and instead worry about the rate of change in the amount of tasks A.I. systems are out performing humans at.
comment in response to
post
And off she goes having understood near zero of what I've actually written...
comment in response to
post
Also its quite obvious from the above rambling you dont know what you're talking about. Waymo are very much using neural nets as the primary algorithm in their cars: it is all about object recognition.
comment in response to
post
Also its model not module
comment in response to
post
Read what I am saying! I am NOT saying we will achieve AGI - it does not matter. All that matters is that the collective set of algorithms we group under A.I. do tasks better than we can. Waymo, AlphaGo, AlphaFold etc etc are showing this.
comment in response to
post
For a start 4o is already an out dated model, we've seen two more LLM models from OpenAI alone that greatly increase the no. of tasks 4o can do. Not to mention text-to-x(video,audio), driverless car models etc from numerous labs. Secondly it does not need to do everything to have a dramatic impact.
comment in response to
post
Doesnt matter whether you think AGI is a scam or not if we create A.I.s that can out perform humans at most economically valuable work that is all that matters. Even 60% would be catastrophic for us. Lee Sedol couldnt believe he was out played by an A.I. you should take note of his story.
comment in response to
post
Couldn't agree more but cant see any other outcome if we create agentic AGI (and we're very close it appears)
comment in response to
post
A.I. will have its own economy (storage, bots, energy, maintenance etc). The problem for us is when that economy out competes ours for resources.
comment in response to
post
Why? It'll make them a shed load of money which is all they want.
comment in response to
post
Doesnt matter how human intelligence works. It only matters in whether it can out perform humans at economically valuable tasks. At the end of the day no matter what you build A.I. will be based on silicon chips and human intelligence on grey matter - they will never be the same.
comment in response to
post
No guardrails... and no revenue. The whole freeware model (which this is - not open source) is based on the idea of cornering the market and then charging or advertising revenue once captured. But if every Tom, Dick and Harry has one then how's that going to work?
comment in response to
post
The only thing our AIs really lack at the moment is the ability to change weights in real time (so called fluid intelligence) and possibly a more advanced visual system.