I’m flabbergasted that anyone disputes this anymore.

▲ jvanderbot 9 hours ago | parent | next [-]

Let me dispel one myth: The bitter lesson can be true, and you can still want a bespoke, customized solution.

If it is true that eventually a general approach that scales with compute/data will supersede a bespoke expert solution, it is not necessarily true that today's expert solution should be thrown away at the first sign of a data-driven model.

It says nothing about what you should use today to build things, it says everything about how you should do research that lasts.

If you're building for billions of users in 10 years, sure go with whatever you think is the bitterest.

But if you're building for right now, you probably mashup whatever works best. That might still be a PID loop, EKF, and Homographies sometimes. But when you hit the frontier of what "works", then you have to reach for new things anyway.

Chesterton's baby/bathwater

WRT humanoids - I am making no statements.

▲ janalsncm 7 hours ago | parent | next [-]

As an analogy, it is currently possible to cross the country in 6 hours.

That fact does not imply that a person in Boston in the year 1850 should wait until commercial flight was available to travel to California.

So the fact that we will eventually have simpler/more efficient algorithms doesn’t tell us anything about what to do today.

	▲	adrianmonk 6 hours ago \| parent [-]
		Or, if you want to make a plastic phone case, 3D printing can absolutely do that for you. But if you look for a way to make thousands of high-quality phone cases quickly and cheaply, you'll realize that injection molding hasn't been made obsolete.

▲ AndrewKemendo 8 hours ago | parent | prev [-]

>It says nothing about what you should use today to build things, it says everything about how you should do research that lasts.

Right! So who are you arguing with lol?

	▲	qlte 8 hours ago \| parent \| next [-]
		"Bitter lesson" gets pattern matched as a stock answer to anything vaguely resembling optimization/compute trends these days. Like, news about internal reorgs at one company that could plausibly be done for any number of boring, bean-counting reasons. From the article it seems more likely OpenAI just isn't prioritizing robotics anymore. Which seems to fit better than abandoning their booth at a robotics expo because "robotics = solved". `The change came recently enough that OpenAI had already registered a booth at CoRL, a top academic robotics conference in Austin, this week. Though the conference app and floor plan marked OpenAI’s spot, when Fortune visited booth #208, it sat empty with OpenAI’s logo nowhere to be seen.`
	▲	ambicapter 7 hours ago \| parent \| prev [-]
		Everyone very usefully saying "bitter lesson lol" in the comments.

▲ ainch 5 hours ago | parent | prev | next [-]

Sutton himself still thinks the current methods are insufficient. The question is not whether the Bitter Lesson is true, but whether we have algorithms that scale sufficiently with compute to warrant no further research.

▲ Yajirobe 8 hours ago | parent | prev | next [-]

DeepSeek disputed it

▲ scottLobster 9 hours ago | parent | prev [-]

The Bitterer Bitter Lesson is that there are no exponentials, only S-Curves.

You can render graphics with the fastest CPU you can find if you like, I'll keep my dedicated GPU

	▲	AndrewKemendo 8 hours ago \| parent [-]
		They don’t conflict