You'll get some hostility around here for all the slop-text, but the board of personas with anti-cheat public attestation seems like the beginning of a useful forecasting tool.
I tried building something similar to make 72-hour predictions about the US war on Iran, but found that the persona subagents were far too naive. They believed official statements and media reports at face value, they failed to read between the lines or apply principles of bounded distrust. They accepted spin and wartime propaganda and didn't give enough weight to underlying incentives. They didn't learn from their mistakes from earlier rounds or downgrade their trust in sources after their statements were repeatedly proven false.
It must be possible to improve accuracy with memory, system prompts, progressively changing subagent weights based on historical performance, etc.
I found it helpful to allow the personas to talk to each other. A pure weight of 20% each for 5 personas that are blind to the arguments of the others didn't work as well as personas that modified their rationales after reading the output of the others.
After each prediction resolves, I would have each persona create a post-mortem analysis of what they got right and wrong. Maybe visibility into prior post mortems of their own persona and those of others on the board could allow them to recognize historical cognitive biases and recalibrate for the next prediction.
Presumably the hosted version will have a leaderboard of some sort. Each board might not be able to cheat, but if users can cheaply create many sockpuppet boards, you'll see the Baltimore Stockbroker Scam emerge on the leaderboard. If the public attestation is to be meaningful, it must be difficult and expensive to create new boards.
No political party is going to say they're "pro censorship".
But if they promise to target specific groups of people to close their schools, regulate how they dress, ban their prayers, and suppress their art, it's all about restricting their freedom of expression.
Örebropartiet policies directly target and restrict the religious, educational, and cultural expression of people who legally reside in Sweden.
Their polices focus on the way people dress, the languages they speak in public, the institutions and schools they build, the traditions they practice.
People would be forced to self-censor their speech, their beliefs, and their behavior.
It should be obvious that it is utterly incompatible with the values and mission of Mullvad for a Mullvad executive to give a large amount money from Mullvad customers to advance public policy with the primary and direct intent to restrict freedom of expression.
>> Being in a tolerant and intellectually open environment ...
Karl Popper said, "Unlimited tolerance must lead to the disappearance of tolerance. If we extend unlimited tolerance even to those who are intolerant, if we are not prepared to defend a tolerant society against the onslaught of the intolerant, then the tolerant will be destroyed, and tolerance with them."
>> the same way that someone's opinions on animal rights, taxes or public healthcare ...
We're not talking about reasonable people disagreeing about tax policy, we're talking about free expression, the entire purpose of Mullvad.
When you make a large donation to a political party whose most fundamental policy is restricting the free expression of people, that is wholly incompatible with everything Mullvad says they stand for.
When a founder and executive with influence over Mullvad policy and operations is exposed actively and financially support restricting free expression of people, it's not "tolerant" to pretend that's somehow compatible with the mission and brand of the company.
Simple minds want to believe one simple thing and then rationalize everything else to force consistency with that one simple idea.
If you want to believe the simple idea that AI is mostly hype, then you'll get stuck in a multi-year loop talking about stochastic parrots, ridiculous valuations, and doomer scaredycats.
But the real world isn't so simple. Multiple seemingly contradictory things can be true at the same time.
Some AI is useless. Some is incredibly powerful and useful even though it makes mistakes. Some companies are wildly overvalued. Some extremely large and expensive companies will quadruple from here. Some frightening scenarios will look silly in hindsight. Other frightening things will happen that none of the doomers foresee.
It would be great to explore those new ideas and possibilities.
It's so boring rehashing the same old tired and worn out ideas like "they're just hyping the danger to pump their shares up."
> programmed to mimick interaction as if it HAD those beliefs and experiences
We spend far too much time debating the essential nature of consciousness when it doesn't matter if it's real (whatever that means) or simulated.
I get far better results in my projects by encouraging the model to argue, to push back, to poke holes in the design, to think creatively about corner cases, to be a devil's advocate, to do lateral web search to find alternatives, to challenge assumptions, to passionately advocate for what it believes is right.
But I don't want to engage all these assholes myself, so I spin them all up as critic subagents with another subagent to listen patiently and be the judge/arbiter.
If I have to choose between sycophancy and assholery, I think assholery gets far better results.
It's a marketplace of ideas where I don't have to suffer through all the unpleasant and overly confident know-it-alls.
I watched the video and I wish I could get those 13 minutes of my life back.
He could have done it in 13 seconds instead of 13 minutes: "Anthropic is lying about the effectiveness of agentic loops because there's this one screen flicker bug in Claude Code that took a year to fix."
Yeah, like when United Airlines claims a plane can fly 300 people 6,000 miles they are lying to you.
I can prove they're lying to you because people have been complaining about uncomfortable seats and flight delays for literally decades and those issues still aren't fixed.
I could do the planning but I don't, for the same reason that I could write the source code but I don't, for the same reason that I could write the machine code but I don't.
They could lock them down legally which would prevent commercial use, but they choose not to, and they boast about how many tens of millions of times Gemma models have been downloaded by developers.
So there must be more to the rationale than just local model weights getting hacked out of devices.
Doesn't that passive process reverse at some point?
The trillions that mechanically and automatically flowed into index funds in pensions and 401k accounts must mechanically and automatically flow right back out after retirement, right?
Especially when younger generations are too poor to save for retirement and most companies don't offer pensions to younger workers any more, where will the inflows come from to offset the outflows?
I tried building something similar to make 72-hour predictions about the US war on Iran, but found that the persona subagents were far too naive. They believed official statements and media reports at face value, they failed to read between the lines or apply principles of bounded distrust. They accepted spin and wartime propaganda and didn't give enough weight to underlying incentives. They didn't learn from their mistakes from earlier rounds or downgrade their trust in sources after their statements were repeatedly proven false.
It must be possible to improve accuracy with memory, system prompts, progressively changing subagent weights based on historical performance, etc.
I found it helpful to allow the personas to talk to each other. A pure weight of 20% each for 5 personas that are blind to the arguments of the others didn't work as well as personas that modified their rationales after reading the output of the others.
After each prediction resolves, I would have each persona create a post-mortem analysis of what they got right and wrong. Maybe visibility into prior post mortems of their own persona and those of others on the board could allow them to recognize historical cognitive biases and recalibrate for the next prediction.
Presumably the hosted version will have a leaderboard of some sort. Each board might not be able to cheat, but if users can cheaply create many sockpuppet boards, you'll see the Baltimore Stockbroker Scam emerge on the leaderboard. If the public attestation is to be meaningful, it must be difficult and expensive to create new boards.