mr international
sf
Joined June 2022
- Tweets3.6K
- Following1.5K
- Followers200
- Likes33.2K
Pinned Tweet
selling ai products is about selling improvement at an increasingly cheaper scale with the simultaneous promise of product improvement.
this goes for OpenAI as much as ANY company selling Agents or AI in any form
that means that as a Startup to win, it’s essential to be providing an ever-increasing amount of value at increasingly high volumes, implying the degradation of company pricing power.
the fact that large companies are buying themselves into this scheme implies that they understand this race to the bottom of that pricing power at an ever increasing pace and low margin.
volume is restricted to some extent now by human to human transaction, need, the only customer is a human for now
the ability for ai and humanoids to transact with one another (even in a separate currency which is only exchanged perhaps into dollars (see: stablecoins) will bring about more ‘customers’ filling in that gap
this is going to feel extremely weird, as the volume of money with humans gets washed with the digital cash
this is the last arc however
for non-decentralized mega-profitability
the debt has piled up and the money wash needs to come before the job squeeze by the rent-seekers
max retweeted
Dopamine does not spike highest for guaranteed rewards. It spikes highest at reward prediction error. If you want to have fun in life, you should stop avoiding risk and uncertainty.
max retweeted
the difference between Grok Voice 3 and GPT-Live-1 is MUCH bigger than the chart suggests.
two big reasons:
- It's single-prompt, while Live-1 requires two different prompts: one for the S2S model, and one for the 'delegated' model (not shown in the $0.05/min sticker price). Want to move your cascade stack (STT-LLM-TTS) to GPT-Live-1 realtime? You'll need two prompts.
- The intelligence with Live-1 isn't in Live-1, it's in the delegated model. That means that by the time your (fast/dumb/cheap OR slow/smart/expensive) delegate model instructs Live-1 where to take the conversation next, Live-1 has already gone two sentences off the rails.
enormous frustration with a multi-week switching cost with still worse-yet instruction following. Grok Voice 3 (with the help of bot) was in production in less than an hour.
and a bonus:
- 11 hours ago, OpenAI announced retirement of GPT 4.1, which was the default LLM of choice for voice agents for nearly all developers in the space. Even as a two year old model, getting any other model to 1) follow instructions, 2) call tools, 3) be as human, and as quickly as GPT 4.1 was nearly impossible.
genuinely game-changing.
max retweeted
Something I told 14 yo: People are going to stop reading books. I wish this wasn't so, but I fear it is. The silver lining in this cloud is that if you're one of the few people who still read, you'll have a huge advantage over everyone else.
max retweeted
The amount of time it takes to assemble a group of friends from scratch in a city you literally know zero people in appears to be about four days if you are willing to embarrass yourself a lot
max retweeted
I know it might kill everyone, but I can't help loving AI. All my life I wanted to talk to an alien intelligence, but I never thought I'd really get the chance to.
max retweeted
In adults, limiting smartphone functionality to texting and calls and blocking all social media and mobile internet for 2 weeks significantly improved attention, self-reported well-being and mental health. 90% of participants experienced a benefit.
not to be forgotten:
"...the incentive for us both as individuals seeking education and as governments providing education a large part of that has been so that we can then get good jobs and contribute to society. So if the vast majority of jobs are being done by machines, then we lose the incentive to become well educated, governments lose the incentive to pay for us to become well educated. And then you kind of break this chain where each generation..."
- Dr. Stuart Russell
(14:02 - 14:40)
youtube.com/live/GrkIMd3A_JA…
So the computers can dream
A "Neural Computer" is built by adapting video generation architectures to train a World Model of an actual computer that can directly simulate a computer interface.
Instead of interacting with a real operating system, these models can take in user actions like keystrokes and mouse clicks alongside previous screen pixels to predict and generate the next video frames.
Trained solely on recorded input and output traces, it successfully learned to render readable text and control a cursor, proving that a neural network can run as its own visual computing environment without a traditional operating system.
arxiv.org/abs/2604.06425
Cool work by @MingchenZhuge @SchmidhuberAI et al.!
“Rapport requires mutual vulnerability, and I cannot be wounded. Without the risk of wound, there's no true north for intimacy.”
max retweeted
Guy who added cross-region replication and failover architecture to his project with 14 users
When faced with a choice choose both
The highest IQ people I know are all either
> fully locked in and building or investing in AI. goal is to work towards generational wealth in the midst of a new technological revolution
> fully checked out of society and the corporate rat race. they are quitting their job, deleting social media and moving to the middle of nowhere to live a quiet life with no distractions
Literally no in between
