Not totally wrong. Self play works well with if your problem can be easily simulated in an RL environment where the model can easily explore different states. RLHF or similar techniques is not that since we don't have exactly have a simulation environment for language modelling
Right now there are companies which hire software devs or data scientists to just solve a bunch of random problems so that they can generate training data for an LLM model. Why would they be in business if self play can work out so well?
> Right now there are companies which hire software devs or data scientists to just solve a bunch of random problems so that they can generate training data for an LLM model.
> language itself is one tool of communication invented to supersede body-language and grunting and noises
That's a pretty utilitarian view of language. How would it feel if everyone spoke and wrote like a PR representative? This is what an article written by an LLM is starting to sound like.
I'm even willing to argue that the way in which you convey your ideas is as important as the idea itself. Like we could all be eating soylent for our daily nutritional requirements but we don't. The taste of the food we eat is important. It's the same with writing for me
I wouldn't make that big of a deal out of it for sure. People already sound that way to me. Everyone saying "have a nice day" when they don't mean it, and don't get me started on office-speak. You don't see me throwing a fit everytime says "let's circle back on this" or "let's take this offline". The LLMs are trained to not ruffle any feathers, similar to office speak. you get that, i get that. the meaning of OP is well communicated. Why are you making a big deal out of it?
I don't think all this crusading has any place in a technical discussion.
Thanks for clarifying this. I was genuinely frustrated with copilot due to the lack of features.
If it's possible please push your large business clients to update office. I work for a multinational pharma company and the copilot feature in excel deployed there is next to useless
Its shocking how they didnt. Imagine how shit the culture must be when employees arent bothering to consider how the user will use the feature, just focussing on getting it through
Here is a fun one. I had a column with around 200 entries and there were some duplicates in it. I just wanted to see which were duplicates and remove some of them.
I selected the cells and asked copilot to tell me which ones were duplicated. Copilot had to ask me to copy and paste the cell contents in its chat box. It couldn't even detect which cells were selected and read them
Why even have copilot inside excel when it can't even read a cell? This is what happens when all you care is about KPI metrics or what not
My recent frustration is with Powerpoint Designer suggestions. Who asked for this? The suggestions don't make sense or look good. And then microsoft provides a helpful tip to not pop open suggestions till the next time powerpoint is opened
My personal take for a long time has been that the primary driver of most war today is boredom. War today is undertaken for entertainment. It's a special kind of entertainment that taps into deep brain stem circuits and provides a false but deeply resonating sense of purpose and meaning. When you hear that "people don't have a sense of meaning," it means their brain stem is not feeling the tribal loyalty emotions connected to warfare.
It would be cheaper to solve resource shortages in almost any other way. I don't really buy that explanation, at least for most wars. I think most wars today have roots that are far less rational.
Note that this applies IMO to all participants on all sides insofar as they had any role in starting or sustaining the war.
I think the primary drivers of war come from the top--powerful people motivated by greed and ego. Those are the spark that starts wars.
Boredom works from the bottom, providing fuel for wars in the form of soldiers. More specifically, young men in particular are easily appealed to by offering them a part in some great heroic endeavor, and a promise to mold them into someone whose manhood and courage may never again be questioned.
Of course, as many former soldiers have found out, you usually receive none of those things. The endeavor was bullshit, you were only a cog, and there is no badge of honor in the world that exempts you from the human experience of being made to feel small.
Wildly disagree with that. I think the overwhelming majority of people want simple, peaceful existence, and that the 'lack of meaning' can be solved through deeper shared community goals and aspirations.
More prominent figures like Trump, Putin or al-Assad don't wage war out of boredom, but out of ego, or visions of a glorious future that only they can impart (which I guess is still ego).
I also think that the various regional conflicts in Africa are in no way driven by the fact that the various political groups are just sitting there with nothing to do.
That said, I do think that a 'common enemy' provides a great deal of focus to communities, as we're wired for it... but the definition of community (who is 'us') is largely malleable and entirely flexible. But it's only one way of providing that meaning.
I also think conflict is largely glorified through American media, which is aggressively pushed on a lot of the English speaking world. The videos of the SF soldiers talking about killing people in Iraq and Afghanistan, and how cool it was with no remorse for the taking of life in a conflict that none of the local population asked for. Of the people I've talked to that have been through armed conflict (specifically Angola, and Serbia), and so strongly against conflict that the reactions are almost scary.
So no, I don't think conflicts are started or sustained out of a sense of boredom.
When one communities deeply shared goals and aspirations conflict with another's (or subgroups) is when you get war and violence. The eras of relative peace is when you have one empire imposing its will.
I think this is skewed by your perception of how frequent wars actually are. If your idea of a typical war is Trump bombing Iran, well, I disagree with your assertion, but it's at least a colorable argument. But those kinds of wars between clearly defined states are actually incredibly rare.
Your typical war, however, looks more like the M23 rebels (backed by Rwanda, though they deny this) fighting the Congo state. Take a more expansive definition of war to include armed conflict in general, and the typical case looks more like the ELN in Colombia. Almost all of these kinds of conflicts can be fairly analyzed as fighting for control of resources, chiefly land and the people or the rents that can be derived from de facto control of that land.
The expanse future isn't that bad - even at the start of the series we've already made it to the asteroid belt and Jupiter moons, and the civilization consists of several sovereign self-governed entities with individual entrepreneurship and private enterprise allowed. It means we didn't annihilate ourself in a nuclear war, nor our civilization collapsed into allways-fully-connected ant colony (or one global fascist/communist/religious regime).
Agreed it’s a tolerable vision, it could be worse. But it’s also a vision of humanity mostly living in enormous disenfranchised structural underclasses - corporate-authoritarianism in the asteroids and subsistence-UBI for all those unnecessary humans on Earth.
It’s a vision of incredible technological progress without any growth in our ability to justly and humanely govern ourselves or move past violent conflict.
I agree with GP this is our current trajectory. I’d live in that world and hope I’d get lucky, but what a disappointment if that’s all we can manage.
I don't know that there was a lot wrong with Earth under the Expanse though.
The problems there were kind of organic: they just didn't need that many people, but they did have UBI, but even if you wanted to better yourself and were exceptional at your job... You could still be 50,001 in the queue of the 50,000 they needed.
Earth in the expanse desperately needed places to expand too and send people, but the solar system just wasn't that habitable.
One of the reasons I love the Expanse so much is how deftly it wove subtle economic and resource dynamics into the plot, while also integrating so many other themes, genres, and styles.
I agree with your analysis of the cause of Earth's troubles, though I'm not sure that adds up to not much being wrong with it. The Earth in the Expanse never figured out how to deal with "excess humans" and the result was planet-Baltimore, that seems pretty wrong to me. And I don't think it's too soon for us to be taking a hard look at how this is likely to work out on the real Earth.
uh I would argue that at the beginning of The Expanse things are middling to bad and at the end things are pretty fucking bad. The epilogue of the final book is the only thing that's unabashedly optimistic.
The main series takes place over about 30 years during which several billion people die system-wide as a result of various wars and terrorist attacks, and uncountably many die in the immediate aftermath of the finale. I love it but it's not really a feel-good story!
Right now there are companies which hire software devs or data scientists to just solve a bunch of random problems so that they can generate training data for an LLM model. Why would they be in business if self play can work out so well?
reply