Hacker Newsnew | past | comments | ask | show | jobs | submit | saimiam's commentslogin

A bad driver never misses an exit.

I don’t understand this story. What lesson/joke am I seeing here?

Raw truth is liberating. The moment people turn it into rigid systems, bureaucracies, political movements, or dogmatic institutions, it gets bogged down by ego, power struggles, hypocrisy, and rules. It starts to serve the evil.

Organizing information could be considered its own form of (self-inflicted) torment.

Ohhh, that’s why Google gave up its “dont be evil” stance!

This sounds roughly at par with Y2K in terms of “you need to fix your shit right now wherever it is” complexity except every day is Y2K for your code base.

(from the trailer) Holmes' mouth reminds me of the Amazon logo. Not just the shape. Also the complete disconnect between the commonly held meaning of a smile and the entity which is currently using it.

Look up Pan Am smile vs Duchenne smile.

Isn’t alignment ultimately an ethics question? Given a strong enough incentive, most humans will also rationalise their shortcuts/crimes. E.g., the hardest part of running a marathon is suppressing the voice in your head telling you that completing the race is pointless.

I’m not at all familiar with AI SOTA but sounds to me that feeding the models enough content about ethical behaviour should help because, as it stands today, these models don’t know how to think morally.

But moral behaviour needs a personality type. Maybe some agents in every cohort of agents need to be told they are like Jesus/Mohammed/etc and see how that influences the behaviour of that cohort?

Even human morality needs guardrails in the form of law enforcement so why should agents be different?


> Isn’t alignment ultimately an ethics question?

It’s a let’s not die question


Who’s “Bobby” mentioned in the article? Did the writer just mix up Billy Evans with Bobby?

Article has been corrected to Billy everywhere Bobby previously was.

Previously: https://archive.ph/sSnek

Corrected: https://archive.ph/k3Hnl


It does seem that way, half-way through they switch from "Billy" to "Bobby"

All Indian banks use bankname.bank.in as their domain. I’m not sure who owns bank.in but this is a common suffix which is different from the .co.uk pattern.

IDRBT Institute for Development and Research in Banking Technology

If something at rest is accelerating at 9.8 m/s^2, how long in seconds will it take to reach 10% of c? Answer to the nearest order of magnitude - will it take approximately 1000, 10k, 100k, 1000k seconds?

I’m sure you know this is an exponential growth question but have no intuition of the answer.


That is a linear growth problem whose answer is very easy to intuit.

It’s not a linear growth problem. Acceleration is quadratic.

As per your intuition, how many seconds later will you hit 10% of c?


Acceleration is not quadratic with respect to speed/velocity, and c is a speed/velocity.

You should have asked something simpler.

Let's say you have infinite ships traveling at light speed originating from earth trying to colonize the entire universe.

All of this is funded by borrowing capital from earth and earth expects a 5% annual return in perpetuity.

The space ships must pay interest to earth and if they fail to pay it, they are not allowed to perform further colonization.

Will earth manage to colonize the entire universe? Aka, can the infinite number of at light speed traveling ships outrun the interest payments?

After answering that question this one should be easy:

What if the universe was infinitely large and infinite growth was possible?


You must be joking. A high schooler with a few hours of physics classes can intuit the answer.

Until special relativity kicks in to completely invalidate whatever intuition you have about this problem.

How is this a joke?

Knowing exponents and how to apply it is not the same as having any intuition about what the actual value of a certain exponential function will be at a certain point and when it crosses a threshold.


There’s no exponential function involved, which is why this is easy to intuit. A true exponential function would indeed be difficult to intuit.

According to UNICEF, globally, Diarrhea causes 444k deaths of children under 5 or ~1200 per day.

Globally, 1.2 million people (all ages) die of diarrhea every year.

India accounts for 120k deaths of children under 4 or roughly one fourth of global D related deaths or 328 deaths per day.

Assuming a similar ratio across all ages, India must account for 300k D related deaths every year of roughly 820 deaths. Not great (obviously) but at least it only accounts for 1/4 of all D deaths, not 1/3rd.

Edit - worth pointing out that India has 1.4B out of 8B humans globally or a little more than 1/6th.

https://data.unicef.org/topic/child-health/diarrhoeal-diseas...

https://www.joghr.org/article/75428-evaluating-india-s-inten...


Such anti patterns are like Nostradamus’s prophecies - you see a company flopping then curve fit a narrative to justify the antipatterns.

How do I, as a founder, know that I am falling into an anti pattern? I don’t believe it is possible to know. You can only “know” in hindsight and that makes these anti patterns useless.


>> How do I, as a founder, know that I am falling into an anti pattern?

The easiest way is to have outside advisors who can and are willingto adversarially challenge the assumptions you're making. If you don't have those, good cofounders can sometimes play the same role (although that is harder since they are likely to share similar biases as you).


This type of advice is much like best practices. One of the core reasons why AI will fail.

Its a bit like cooking advice

- Products should be fresh

- Use the best ingredients

- Salt early and correctly

- Cook by temperature, not time

And so on but then you cook side by side with a pro chef...and does not taste the same...


or worse, what if an anti-pattern is "your" way to success

It takes understanding your own impulses and cognitive biases.

https://en.wikipedia.org/wiki/List_of_cognitive_biases

Most people start a small business for tax reasons, and the few lucky loud clowns claiming hyper-scaling BS is sustainable simply disappear within 5 years.

Best advice, answer the simple question "will random people buy this when no one is looking?"... if you have to explain/sell a product/service more than 5 seconds, than that business model likely isn't sustainable.

I also see a lot of smart folks that think they will get paid for excuses... only lawyers, PR, and politicians are in that business. =3


> Best advice, answer the simple question "will random people buy this when no one is looking?"... if you have to explain/sell a product/service more than 5 seconds, than that business model likely isn't sustainable.

For B2c, maybe. B2B is very different - they aren't going to make up their mind in 5 seconds and you aren't going to have a product that can be explained in 5s.


>B2B is very different - they aren't going to make up their mind in 5 seconds

If not, than one is explaining/selling again, and you will only ever have 2.7%*(time/sales_pitch_time) customers. Letting client/customers dictate the terms of a product/service often will just turn your business into a staffing agency.

B2B is actually more risk, as any deal exceeding 12% of revenue sets a timer on how long your business will likely survive. Some unlucky folks also end up acting as a product warehouse for clients, contract labor, or a scapegoat for client internal strategic mistakes.

If people can't summarize what they do in 1 sentence, than they likely have a fuzzy definition for a business project goal. For example, people that use more than 4 conditional statements (if/but/when etc.) to describe something, than I have observed they often last less than 3 years regardless of cash burn rates. =3


> If people can't summarize what they do in 1 sentence, than they likely have a fuzzy definition for a business project goal.

I broadly agree with your point, but disagree with this sentence. IME a 1-sentence business goal is almost certainly too fuzzy to be useful. There's nothing actionable about "We help you survive audits and readiness for regulation $FOO", which is frequently the business goal.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: