HNHacker News
TopNewBestAskShowJobs

einrealist

1,436 karma · joined December 16, 2013

Former CTO at a startup. Now a senior consultant at Evil Corp, again.
submissionscomments
einrealist··on Dispelling misconceptions about RLHF
A World Model is a theoretical type of model that has knowledge about the "real world" (or whatever world or bounds you define). It can infer causalities from concepts within this world.

Yes, you can use Genie 3 to train other models. Its far from perfect. You still need to train Genie 3. And its training and outputs must be useful in the context of what you want to train other models with. That's a paradox. The feedback loop needs to produce useful results. And Genie 3 can still hallucinate or produce implausible responses. Symbolism is a wide term. But a "World Model" needs it to make sense between concepts (e.g. Ontologies or the relation of movement and gravity).

einrealist··on Dispelling misconceptions about RLHF
> “Successful” is importantly distinct from “correct.”

This is the most important sentence describing the fundamental issue that LLMs have. This severely limits the technology's useful applications. Yet OpenAI and others constantly lie about it.

The article very clearly explains why models won't be able to generalise unless RL is performed constantly. But that's not scalable, has other problems in itself. For example, it still runs into paradoxes where the training mechanism has to know the answer in order to formulate the question. (This is precisely where the concept of World Models comes in or why symbolism becomes more important.)

LLMs perform well in highly specialised scenarios with a well-defined and well-known problem space. It's probably possible to increase accuracy and correctness by using lots of interconnected models that can perform RL with each other. Again, this raises questions of scale and feasibility. But I think our brains (together with the other organs) work this way.

einrealist··on Why LLMs can't really build software
> Twitter sucked [...] Electric cars sucked [...] Phones sucked

All these things are not black boxes and they are mostly deterministic. Based on the inputs, you EXACTLY know what to expect as output.

That's not the case with LLMs, how they are trained and how they work internally.

We certainly get a better understanding on how to adjust the inputs so we get a desired output. But that's far from guaranteed at the same level as the examples you mentioned.

That's a fundamental problem with LLMs. And you can see that in how industry actors are building solutions around that problem. Reasoning (chain-of-thought) is basically a band-aid to narrow a decision tree, because the LLM does not really "reason" about anything. And the results only get better with more training data. We literally have to brute-force useful results by throwing more compute and memory at the problem (and destroying the environment and climate by doing so).

The stagnation of recent model releases does not look good for this technology.

einrealist··on Debt of AI Knowledge
That's why I call it "Vibe Prompting" instead of "Vibe Coding". If the model is changed by the gate keeper, your previous prompts might just be useless.
einrealist··on Trump plans 100% tariff on semiconductors, with big exemption
Good observation. To Trump, tariffs are a simple tool that he can use to demonstrate that he is taking action. It's a spectacle, something to make him appear busy. He is not interested in the details or the actual effects. There is no strategy. Everything is just vibes. He is satisfied by vague, non-binding promises. Trump is inherently stupid and lazy.
einrealist··on Tim Cook presents President Trump with award
I am not certain whether Trump recognizes how absurd this looks. People are playing him and his narcissism like a fiddle. Trump is too stupid to realize it. And he has no shame.
einrealist··on Poorest US workers hit hardest by slowing wage growth
That's the myth of Trickle-down economics in action.
einrealist··on OpenCQRS – an open-source CQRS framework for the JVM
Interesting. Looks much simpler than Axon. I like the framework paradigms, how it is wired and how simple the domain classes can look like.
einrealist··on I've Had It with Microsoft
Same here! I did cancel my subscription. I hadn't used it much aside from Outlook anyway. I probably wouldn't have cared about it if there hadn't been a price increase. Its now Thunderbird and LibreOffice.
einrealist··on Intel CEO Letter to Employees
Will be answered by the ASI - soon. /s

But to be serious: Intel creates software too (compilers for example). Nvidia does provide software too. Its also nice to sell licenses.

einrealist··on Two narratives about AI
It's hard "to tune out" from the topic when investments in LLM technology surpasses revenues of nation states, and when US policy is not trying to prevent harm to society, but making it more likely. By 'harm', I don't mean some 'rogue AI' sci-fi scenario. I mean a financial time bomb that far exceeds the 2008 subprime mortgage crisis.
einrealist··on CARA – High precision robot dog using rope
Aside from the onboard computing required to power the AI model controlling the robot, which also consumes energy, it is important to be as efficient as possible. However, if an Optimus is intended to replace human labour, such as lifting objects and wander around with it, the energy efficiency of the platform becomes negligible.

The platform in the video and the robot dogs from Boston Dynamics are ideal for tasks where they are only limited by their own weight and the amount of computing power required to navigate, such as exploration.

I suppose that's why we mostly see autonomous delivery robots on wheels.

Or maybe I am being too pessimistic about other platforms...

einrealist··on AI overviews cause massive drop in search clicks
At what point does this remove the 'platform provider' defense for wrong or defamatory information generated by Google & Co?
einrealist··on AI overviews cause massive drop in search clicks
This raises the question of when it becomes harmful. At what point would your company issue a cease-and-desist letter to Google?

The liability question also extends to defamation. Google is no longer just an arbiter of information. They create information themselves. They cannot simply rely on a 'platform provider' defence anymore.

einrealist··on CARA – High precision robot dog using rope
What is the power consumption of these robots? I often wonder how limited and viable autonomous robots really are. When I look at Tesla's Optimus or Boston Dynamics' spectacular robots, how quickly do they need to be recharged?
einrealist··on Starship was doomed from the beginning
I wonder by how much Musk and Bezos are driven by science fiction and how this materializes in the real world. Its okay to have dreams and goals. However, as soon as real resources (as in energy, materials, environmental impact) are spent, we better be sure that things are viable from a grounded scientific perspective.

For example, Musk's obsession with colonising on Mars is troubling when you hear what scientists have to say about it. Musk tends to talk about it as if Earth is already lost.

So, how many decisions by the wealthy and gullible (or downright stupid, orange) politicians are driven by science fiction and vibes?

einrealist··on AI agent benchmarks are broken
And boil the planet.
einrealist··on Grok 4 Launch [video]
So the progress is basically to brute force even more?

We got from "single prompt, single output", to reasoning (simple brute-forcing) and now to multiple parallel instances of reasoning (distributed brute-forcing)?

No wonder the prices are increasing and capacity is more limited.

Impressive. /s

einrealist··on The force-feeding of AI features on an unwilling public
Your "API credits" don't buy the model. You just buy some resource to use the model that is running somewhere else.
einrealist··on The force-feeding of AI features on an unwilling public
Isn't "Engineering" is based on predictability, on repeatability?

LLMs are not very predictable. And that's not just true for the output. Each change to the model impacts how it parses and computes the input. For someone claiming to be a "Prompt Engineer", this cannot work. There are so many variables that are simply unknown to the casual user: training methods, the training set, biases, ...

If I get the feeling I am creating good prompts for Gemini 2.5 Pro, the next version might render those prompts useless. And that might get even worse with dynamic, "self-improving" models.

So when we talk about "Vibe coding", aren't we just doing "Vibe prompting", too?

einrealist··on The force-feeding of AI features on an unwilling public
If those smaller models are sufficient for your use cases, go for it. But for how much longer will companies release smaller models for free? They invested so much. They have to recoup that money. Much will depend on investor pressure and the financial environment (tax deductions etc).

Open Source endeavors will have a hard time to bear the resources to train models that are competitive. Maybe we will see larger cooperatives, like a Apache Software Foundation for ML?

einrealist··on The force-feeding of AI features on an unwilling public
The major AI gatekeepers, with their powerful models, are already experiencing capacity and scale issues. This won't change unless the underlying technology (LLMs) undergoes a fundamental shift. As more and more things become AI-enabled, how dependent will we be on these gatekeepers and their computing capacity? And how much will they charge us for prioritised access to these resources? And we haven't really gotten to the wearable devices stage yet.

Also, everyone who requires these sophisticated models now needs to send everything to the gatekeepers. You could argue that we already send a lot of data to public clouds. However, there was no economically viable way for cloud vendors to read, interpret, and reuse my data — my intellectual property and private information. With more and more companies forcing AI capabilities on us, it's often unclear who runs those models and who receives the data and what is really happening to the data.

This aggregation of power and centralisation of data worries me as much as the shortcomings of LLMs. The technology is still not accurate enough. But we want it to be accurate because we are lazy. So I fear that we will end up with many things of diminished quality in favour of cheaper operating costs — time will tell.

einrealist··on U.S. bombs Iranian nuclear sites
Those UN buildings in Geneva and New York, are they up for sale already? What about the buildings of the US Congress?
einrealist··on Is there a half-life for the success rates of AI agents?
So as the space for possible decisions increases, it increases the likelihood of models to end up with bad "decisions". And what is the correlation between the increase in "survival rate" and the increase in model parameters, compute power and memory (context)?
einrealist··on Accumulation of cognitive debt when using an AI assistant for essay writing task
"I"? You should treat yourself as an anecdotal exception.

You are reading on HN. You are probably more aware about the advantages and shortcomings of LLMs. You are not a casual user. And that's the problem with our echo chamber here.

einrealist··on Salesforce study finds LLM agents flunk CRM and confidentiality tests
Remember that increasing the accuracy/correctness does not solve the problem. It only increases the cost of identifying cases where the LLM has failed.

That's why I am highly sceptical about using LLMs in situations where accuracy matters. And that's even if humans are kept in the loop (we are lazy and are biased towards trusting computations).

einrealist··on Tesla's Optimus and the Humanoid Robot Race Nobody's Ready For
Found a TSLA shareholder.
einrealist··on Telegram, the FSB, and the Man in the Middle
For this reason, I never even considered using Telegram. If I were an unethical intelligence service like a Russian one, I would create a messenger app (and/or social network). Based outside my jurisdiction would add to plausible deniability.

On another note: I wonder how many of those VPN services are actually fronts of intelligence services.

einrealist··on The Illusion of Thinking: Understanding the Limitations of Reasoning LLMs [pdf]
All that to keep the investment pyramid schemes going.
einrealist··on My AI skeptic friends are all nuts
Exactly. And even if LLMs become more accurate and qualitative better, it just raises the costs for validation. Errors and bad design decisions become harder to spot.
← PreviousPage 4 of 19Next →