The Myth of the Objective
opowell.com
opowell.com
His idea is you should focus on the journey (always improve) rather than artificial endpoints. It's a profound idea, and one I've sometimes struggled to come to grips with.
[0] http://asq.org/learn-about-quality/total-quality-management/...
Going by his theory... If you say, "I am paying you to produce 1,000 cars per month" then that's what you'll get. If instead you rely on intrinsic motivation, and improve the system, you will get more than 1,000, and they'll be higher quality.
This doesn't work everywhere. (Certainly not with Sales commission!) But it is worth thinking through the logical implications of it.
It’s too bad these methodologies aren’t more prevalent in corporate culture. Imperative Command & Control methods always seems to dominate.
I really should’ve known that already :)
It's apparent that it's actually saying that being too "greedy" (i.e too low a horizon) can be bad for achieving an objective. This is well known to both Control theorists, RL practitioners, and many other non-STEM people (in marketing/politics/sociology etc.).
This triviality has nothing at all to do with Zen or any of the things Westerners take for being "rustic" and "warm and fuzzy".
It's basically just talking about the incredible power of broadly pursuing novelty and interestingness, as opposed to a specific goal.
Another phrase for it would be "unknown unknowns": if your success criteria are too focused, you'll miss out on solutions that you couldn't conceive of before they're discovered. The "Objective" criticised in this article is like a "known unknown".
Ah, but you see when Plato was saying pharmakon (drug, poison) what he actually meant was pharmakos (scapegoat); since he didn't actually say pharmakos, by its absence that must be the true meaning!
An example: You plan to be a millionaire, and you currently have nothing in the bank. How do you get there? Any plan would most probably not work, if executed in a rigid manner (or we would all look up the recipe on the internet and be millionaires). You end up trying all kinds of jobs, investments, take opportunities when they arise, etc.. and finally you realize that you probably will not ever make it. Or you get lucky and make your million - who knows. Picbreeder showed that it's quite rare to hit the very desired result you wanted.
However, if you were more flexible regarding your goal, at every step, then you might take the job that agrees the most with you, meet a lot of great people, travel if you are able, start a family.. and end up without a million in the bank but with a satisfying life situation. With pic breeder, certain shapes and beautiful pictures emerged that were not intended or planned from the start. The algorithm went along with whatever was there, and optimized on that.
Looking at the millionaire example again - if you already had 500k, then doubling that might not be that hard (?), or not that many steps away. People keep saying that the more money you have, the easier it is to earn - think opportunities like buying an apartment in a firesale and reselling it for 1.5 as much.
For me, it's a very counter intuitive way of looking at things. It goes very deep against the wanting of goals or things in our current society. OTOH, it somehow makes sense that things work this way, because every step in our step-by-step plans is SO dependent on things beyond our control. Agile development methods remind me of this very much: Try telling a customer not familiar with agile that the software she orders might look completely different than she imagines it right now - most will think you are crazy or want to cheat them. But you might recognize problems in every sprint, and adapt the end goal accordingly.
To me, one of the big takeaways here is that it's important not to be so hyper-focused on the local problem space that you overlook potentially better solutions and end up at a local optimum to your overall detriment.
As for the rest of the article, it's mostly hand-wavy garbage.
Most people who are millionaires don't get there by holding a million dollars in currency, they get there by holding assets worth a million dollars. This is not zero sum - assets can be created (and destroyed), and their value is assigned only at the time of transaction. The total value of all assets in the U.S. is significantly larger (by orders of magnitude) than the total amount of U.S. dollars available.
Imagine I create a new shoe factory right next door to an existing shoe factory. After 10 years we find...
1) I failed, I lost all the investors money and had to close down. I destroyed the investors wealth.
2) I win, the next door factory has closed down and now I have all of their business. But I have exactly the same costs, sales, staff and so forth. So the profits come to me from now on instead of the neighbor. I have supplanted them, but consumers and the economy as a whole is no better off. Wealth is transferred to me as ongoing profits that would otherwise have gone next door.
3) I win, the next door factory has closed down and now I have all of their business. I produce the same output as they did but with less staff and lower costs. So consumers benefit from lower prices, the money they saved can be spent on other things, creating demand elsewhere. Wealth is transferred to me as ongoing profits that would otherwise have gone next door. Everyone wins except I employ less people than the neighbor so there is less employment. But the extra demand for other goods because of my lower prices means they get employment at other businesses!
Number 3 is why the UK, which used to employ 90% of people in agriculture before the industrial revolution, does not have 90% unemployment today, the people released from agriculture because of better productivity are freed to work in other sectors instead.
Isn't there also at least a fourth and firth scenario, one where the new shoe factory wins but has larger costs than the old one (better marketing but worse cost control), and another scenario where both shoe factories thrive? What about one where they merge, or a holding company buys both and operates them both to produce different lines.
For any startup that subsidizes its products below its costs it is possible for it to destroy value. If they produce something for $10 dollars then sell it for $8 to someone who derives $9 of value out of it, then they have transferred $1 of wealth to their customer and have destroyed $1 of wealth.
>All goods in circulation in the U.S. have a finite value, and it definitely is a zero sum game.
Except I can create new goods, possibly for free (e.g. writing software, growing crops, mining materials) and directly increase the value of the economy. Contrary to zero sum.
>You can add more goods and therefore value, but that can only happen at a certain rate
Unless you want to try to argue that the rate is so low that the economy is *effectively zero sum in the short term, you've just contradicted yourself. If you can add value to the economy, it is fundamentally not zero sum.
You seem to be conflating the finiteness of wealth and value with the inability to create new wealth or value.
Now, that aside, one could argue that currency exchange itself is zero sum on short timeframes when new money is not printed, but market forces dynamically assign value to currency, such that the economy may still grow with a finite supply of money. Furthermore, I'd like to point out that generally when one purchases goods or services, even in the short term the transaction is unlikely to be zero sum, because goods and services can be used immediately to generate more wealth, and therefore are arguably worth more following the exchange.
Something else to consider is that in your example of "for free" wealth creation you're not actually getting any of that for free, that's a form of wealth transformation or transfer. Let me elaborate on that point using each of your examples. I'll start with mining as that's the simplest, in that case you're taking a natural resource (which is finite) and extracting it and refining it. You're having to pay your workers (and/or buy and maintain machines) in order to do so, so in part your redistributing the companies wealth to the workers and service providers your company does business with. In exchange you receive raw and/or processed minerals/metals. That might seem like wealth creation but it's really transformation, you've reduced the value of the land you extracted the material from and converted it into a transportable form. The value of that material might seem to be more, but that's only because you've invested value in extracting it, in other words you're passing on your cost of doing business. You haven't added value, the value was already there, you've simply converted it and invested some of your companies value into it, so when you sell it you're simply converting one form of wealth into another, you're converting the wealth of that processed material into cash wealth.
The situation with growing crops is similar, although part of what is being invested there is time. You might think, "well, time is infinite, there's always more time", but each persons time is finite and it has value, even if only to that person (opportunity cost), so once again you're doing a wealth transformation, you're transforming those workers time into money, and ground, seed, fertilizer, water, and sun into crops. When you sell those crops you're once again recouping your cost of doing business. Wealth hasn't been created from nothing, it was transferred and concentrated from a variety of sources. You might think, "well, what about the workers time, that's new wealth", only it isn't, there was a cost involved in those workers upbringing and living, so that's once again just a form of wealth transfer and transformation. Truly the only free wealth in the entire thing is the sun, although even that isn't infinite, even if it is free from the perspective of anyone on Earth (it might be more accurate to say it's wasted/destroyed if you don't use it). Ultimately there is no free lunch, entropy always wins.
Software is the most complicated one, as there's zero unit costs associated with it, but substantial development costs. Once again though, you're looking at wealth transformation/transfer. In the case of software you're transferring/transforming the developers, QA, and other workers personal knowledge and time into software. Similar to the workers in the previous example they've invested time and money into improving their knowledge and living, so you're really paying them for them to recoup their losses (wealth transfer) and then when you sell the software you're simply passing those expenses on to your customers.
Ultimately it's all about wealth transformation and transfer. There is finite natural wealth, it existed before humanity, and if humans vanished tomorrow it would continue to exist. Economies are mostly about taking the existing wealth and distributing, concentrating, and transforming it into forms that are more convenient for people. When you get down to it, the unit cost of a good is really it's intrinsic value, it's a form of wealth transformation. Profit margins on the other hand, are wealth transfer, you're transferring wealth from the purchaser to yourself. No value is actually being created. New wealth only comes from discovering new resources. Want to create wealth now? Do like Elon Musk and others are doing and take a look at asteroid mining.
As in Reinforcement Learning, it's an exploration vs exploitation tradeoff.
After all, all the comments about the folly of greed apply just as well to choosing moves in Go, but you don't need to abandon 'the myth of winning' to win Go. It's simply about having an excellent heuristic for promising directions and keeping options open.
However, I can't help but feel as if some of these statements aren't very concretely true, or untrue.
For example, ok, if I get tunnel vision and just focus on one objective, I might lose sight of other paths.
I can see what you mean, but let's put that into practice.
"If you focus on profit, you might lose focus on innovation."
Ok, but if I focus solely on innovation, what happens when I'm not selling anything, or selling things at a loss?
The bills come due at some point, and so many people fail.
We don't all have endless sums of money that we can rely on while we wait for our Zen powers to coalesce and then people magically send us money.
Bill Walsh's The Score Takes Care of Itself is a fun book that, as the title suggests, deals with this idea.
key word for me is *can. Objectives are pretty good in some cases, but not all. Or perhaps, when they ran their machine learning, they needed to first run some unsupervised learning =)
Peter Thiel gave him money and he demolished Myspace somehow.
I don't think he got there by daydreaming ^.^
The central theme is regarding how it is better to chase goals obliquely rather then head on, especially when goals correspond to holistic/unspecifiable things like "wellness". He has case studies of companies that managed to innovate or not, depending on how carried away they got with their objective.
A dead end only means you're looking at the wrong direction at _that_ moment.
Always chase the yet unlearned.
There are AI projects that attempt to approximate a "natural image manifold" – essentially a blob in the space of possible images where you find things that look like photographs to humans. I think the PicBreeder thing has more to do with that; the set of images that people want to create occupies a relatively confined space within the set of all possible PicBreeder outputs. So working off of any previously built image is basically a "shortcut to the manifold", if you will.
Or to put it another way: when you consider the entire space of PicBreeder outputs, any existing image is much closer to whatever you want to make than starting from scratch.
> Many natural processes don’t appear to be objective based. Natural evolution and human innovation are cases in point.
Is not correct. Evolution is highly objective based and micro-incremental.
Regarding evolution, having read both Kenneth's book and "The Blind Watchmaker" by Dawkins I can tell you that Kenneth rediscovers what Dawkins explained in his book in 1986. The experiment Dawkins describes in his book resembles the picbreeder experiment a lot.
According to Dawkins, Evolution has two objectives. Optimizing survival and procreation. We saw how the randomness toward an objective failed to produce anything meaningful in this example so how is this similar to natural evolution?
I see no similarities.
Evolution has two eyes on the path (optimizing survival and procreation) and none on objectives (concrete outcomes, i.e. specific organisms).
> Objectives can, ironically, be obstacles to innovation and creativity.
This, like much of the Zen worldview, is self-refuting. If objectives can be obstacles that leave me with only one eye on the path, then the new objective is just to focus on "the path" of innovation and creativity. My objective may have changed, but I still have an objective.
They don't intend to teach you anything. Instead, they create doubt in your beliefs so that you can see possibilities you were previously blind to.
In this instance perhaps the difference is using the "objective" as a starting point rather than a desired end result.
Most of the claims of the video and article seem to lead nowhere, or are commonly known and while mildly interesting, are easily explained by logical observation and analysis. Knowing them doesn't help get you anywhere (that would be an objective). If there is no objective to the video, then why watch it. This is starting to sound like some kind of relativism.
The success of PicBreeder is accomplished by individuals ability to see interesting things evolving and the vast number of real world similar objects they had to choose from. Of course an algorithm cannot replicate that because it is not creative. This does more to dis-prove natural evolution IMO because the images that took shape required creative intelligence to facilitate them doing so. As to the point of the final pictures only appearing when there was no particular objective, OK fair enough, but how does this help us? What new thing did we learn? We know that many inventions and innovations were discovered this way. Why is it an epiphany? The computer could not create a recognizable picture because it was only given one. If it had the google images library to match to, it would certainly come up with some very great pictures.
On the Living Image Project, The recognizable images didn't happen, because different people saw different things and voted different directions. This failed precisely because there was no objective. If the Living Image Project stated an objective (A teacup for example), it would have had far more success. This just goes to show that objectives are important because in the real world, there are needs. If the article and video was useful, it would have revealed something helpful about how to fill needs (objectives) in a better way. The majority of problems are solved with objectives in mind. This article and video never seem to recognize that. The conclusion almost seems to say that attempting to fill known needs is a problem in the world today.
In the Novelty search demo that was shown, the biped that used the novel behavior algorithm walked better than the farther distance algorithm only explains the constraints that the failing algorithms had. All it proves is that when an algorithm has more choices to explore, in many cases it did better. To me this would be like telling someone to build the best house, car, ETC but only use the knowledge from 3 books. Tell someone else to do the same but use the internet, unlimited books, Talk to people ETC. Should we be surprised at the result? Kenneth Stanley seemed to be, and seemed to think it revolutionary. What new thing does this teach us in practical terms? Nothing. More stepping stones mean a better result. This is common sense.
It is no wonder the National Science Foundation would not give this guy any money. He hasn't taught us anything new. I want my 30 minutes of life back.