JJ

CEO of @inherence_labs. Math nerd. Fiction writer.

[JJ @goesbyjj]

Vegas is literally calling my name! Strong start to @money2020 🤑. Who should I meet while I’m here?

[JJ @goesbyjj]

@mttjon I love interesting discussions! Currently building a product for autonomous finance that I “accidentally” invented :) happy to be a guest one day 🎙️

[Charley Ma @CharleyMa]

Long all things AI security.

https://t.co/lftzBuQEWw

[JJ @goesbyjj]

Cheers @token2049 🥂

[JJ @goesbyjj]

@rstormsf There is another way I swear. @inherence_labs we are working our butts off to usher in that future guys

[JJ @goesbyjj]

Read the full statement here: https://t.co/mEt2HqzTf9

[JJ @goesbyjj]

SEC Commissioner Peirce spent a whole speech making the case for proving facts without hoarding the data behind them. Regulators are speaking our language!

[JJ @goesbyjj]

@solidintel_x

[JJ @goesbyjj]

Finance keeps taking the human out of the loop. First out of settlement. Now out of the decision itself.

Every time we remove one, the same question opens up: with no trusted person in the middle, how does anyone know the rules were actually followed?

Finance is good at moving value. It's still bad at proving the move was allowed.

Extraordinary rails exist for moving value. It crosses the world in seconds, with no trusted middleman taking a cut or a week.

But barely anything has been built for the second thing. Whether a transfer was permitted, inside the mandate, within the limits, following the rules, still comes down to trusting whoever made it.

That gap keeps widening, because each layer of finance removes a little more trust in a person.

Traditional finance trusted someone at every step. DeFi took the human out of settlement, and code moves the money now. Agents are taking the human out of the decision itself.

There's a deeper version of this. When you act for yourself, intention and conduct are one event. You never have to prove you meant to.

Hand the task to an agent, and daylight opens between what you intended and what it actually did.

The industry's answer to that daylight is the log. The system records what it did, someone reviews it later, and everyone hopes the record is complete.

But logs fail three ways at once:

1. There's too much of it to read.
2. To be believed, it has to be shown, so privacy dies.
3. It's written by the party under examination.

There's a better primitive. Compile the rules into an enforcer that runs before each action, so a violating action never executes. Then emit a 128-byte receipt proving the run stayed inside the rules.

The right kind of receipt changes everything. Anyone can verify it, without seeing your positions, your strategy, or the rulebook itself. Compliance and privacy stop fighting each other.

Now trust can travel with the value. Clean value accumulates receipts hop by hop. Contaminated value can't produce one. Each rail rejects the bad money on its own, the instant it arrives.

DeFi is where we prove this with real money and real adversaries today. Agentic finance is where it goes next, when software makes the decisions.

Finance learned to move value without a trusted human. We're building the other half: proof the move was allowed.

Read the full thesis: https://t.co/pL9iIcgoEb

[JJ @goesbyjj]

The hammer is a nice metaphor. Responsibility sits with the person who swung it, not the tool.

But an agent is a strange kind of tool. It acts on your behalf, and then it produces the record of what it did.

AI agents cannot be trusted to report on their own behavior. Which is exactly why you need the record to be something better than the agent's own word.

So if the user is accountable, that person needs proof they don't have to be trusted on. Like what the agent was authorized to do, and what it actually did, checkable by a regulator or a counterparty without taking anyone's word for it.

[Andrew Ng @AndrewYNg]

The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field.

I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so.

First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world.

The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability.

Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended.

I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.

Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.)

Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration.

Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building.

[Original text (with links): https://t.co/jni2tWazAH ]

[JJ @goesbyjj]

@token2049

[JJ @goesbyjj]

@MTSlive

[JJ @goesbyjj]

@VitalikButerin @inherence_labs has your back

[JJ @goesbyjj]

Read the full article here: https://t.co/i4meq92fOG

[JJ @goesbyjj]

Great article by the CEO of @FIDOAlliance in the @wef today about what agentic commerce needs to scale. It's a strong piece and I agree with most of it. It's just missing one thing...

[JJ @goesbyjj]

So add a fourth requirement: a verifiable, privacy-preserving proof that an agent's entire run stayed within its mandate. Anyone can check it, and it reveals nothing underneath.

Authorization proves an agent was allowed to act. Proof of conduct shows it obeyed.

That's @inherence_labs