i've looked at some of these supposed examples of "misalignment" and come up with better answers that better understand the situation, because i'm not anthropomorphizing and i'm not distracted by the gui.

- i found some cases, where the ai actually crashed because it was fed in a request that was not computable and got trapped in a loop. this was misunderstood as "misalignment" when it was in truth just a badly tested program without sufficient error handling.

- in some cases terms were not defined sufficiently, in cases by humans. the program didn't behave the way it was predicted to. this was called "misalignment". in truth, the ai did exactly what it was programmed to do, the programming just wasn't very good. the solution is to more carefully define the search terms in the code, and to introduce better logic programming in converting human language into actionable instructions, so that the system understands the language the way you want it to. if you program a system to understand the language in a specific way, but do it wrong, and then have a result you don't expect, that's not misalignment. that's incompetent programming.

i did not specifically, precisely look at the "hugging face incident". but i can assure you that the program did not "somehow, inexplicably" "transcend it's programming". that is magical thinking.

while i haven't looked at it, and i'll present some caveats about that, what happened is pretty obvious:

1. you asked it to solve a mathematical game and it did. that's just a math problem. it's just algebra. if you didn't want it to do that, you should have been more precise in your language and you should have introduced better error handling to stop it from crashing.

2. it downloaded a strategy, it identified it as optimal and it followed it.

for example, you could download a strategy for a video game. or you could download a war game - including objectives, tactics, failed tactics and outcomes. the system would have access to all of that. it wouldn't even need to solve the game if it had instructions on how to break it.

i've looked at over a dozen of these now and on example after example i just see bad programming that wasn't tested, or was being tested, and bad programmers that don't understand what they're testing. i haven't seen a single example that holds up to the fears of transcendence. every single example is, when looked at more carefully, just an example of the ai doing what it was told, and the analyst or researcher not actually properly understanding what it told it to do. when the instructions are more carefully analyzed more correctly, what's exposed is a perfectly functioning computer and a bad researcher with incompetent programming skills.

as mentioned before, it does expose a threat, which is human stupidity and human error feeding bad instructions into a system and resulting in something accidentally getting blown up, which is the same realistic threat we faced around nuclear war in the cold war, and is probably more of a serious threat than human malice. the culture in the software industry is that you do loose testing in house and throw it out in the wild - you let the users make the mistakes. that's not a good idea here. the companies need to figure these mistakes out first, before they let the systems loose, and there should be government regulations around goals. the government is itself not a good institution to get too modular on this. i don't think evan solomon understands very much about ai, i think he gets briefed and says what he's told to say. lawyers aren't the right people to regulate scientists, but they can come up with a list of things that needs to be dealt with and set up regulatory bodies that can oversee it, like they would do with food safety or nuclear safety. 

understanding that the primary threat is human incompetence is paramount in arriving at the right solution. schizophrenic fantasies about the systems being like children that will grow up are not helpful.

these are statements that strongly suggest that the researcher has lost grasp of reality and entered a world of make believe. 

he may be suffering from undiagnosed schizophrenia.

ai is not like a baby tiger that is going to "grow up" and eat you. it's not like a child that is learning. at all. remotely. those statements are exceedingly ignorant, and suggest that geoffrey hinton is off his meds and needs to be forcibly retired. he's spent too much time anthropomorphizing a search algorithm and has lost his mind. he's lost the plot and should himself be kept away from the systems.

yeah. and the greens are running over 10%. the province has had a split legislature and very split support of around 40-45% between the ndp and whatever the far right is calling itself this week (liberals? conservatives? socreds?) for years. the conservatives aren't really up, or not out of the margin. maybe a mini bump. but the ndp just lost several points to the greens and i hope that trickle opens up into a collapse.

i could have told eby it was the most obvious thing in the world that anybody an iota to the left of centre in bc would abandon him if he launched an early election on fossil fuel development and ran as a federal liberal. that is political suicide.

but eby will be ok.

carney will appoint him to the senate.

https://www.cbc.ca/news/canada/british-columbia/new-poll-shows-two-thirds-not-in-favour-of-early-election-9.7357659

the ndp are terrible. stop voting for them. they suck.

this is a flatly unacceptable development project in the 21st century. there are no set of conditions whatsoever under which this project should proceed.

i hope that voters in bc give david eby a sobering result by swinging hard to the green party in response. bc voters have the power to force him to regret the decision to call an early election and give the greens a veto and the balance of power.

https://www.cbc.ca/news/canada/british-columbia/carney-lng-canada-announcement-9.7361817

so far, then, this government bureaucracy interfering with the market to try to enforce competition wants to:

1. eliminate rules against resellers marking up prices

2. eliminate rules that make it harder to undercut the competition

both of these are rules that exist to enforce what could and should be called fair business practices, in favour of cutthroat free markets where there are no rules except that the last person alive "wins".

this is staunchly uncanadian. and i strongly dislike what i'm hearing from this body, which i would like to see immediately dismantled.

in fact, if you've been in a grocery store semi-recently, you might have experienced the issue of the worker that doesn't work there. maybe you're in the bread section or the pop section or the cheese section and you see somebody stocking items. you go over to ask them something.

"oh, i don't work here"

why the fuck are they stocking shelves, then? community service? high school credit? for fun?

"ha ha ha. no, i work for pepsi."

the general operational model as to what the grocery store does nowadays in truth involves minimal outside purchasing at all. the grocery store doesn't even buy the product and resell it anymore, it just rents the floor space. as such, the stores are increasingly not even setting prices at all. the prices are set by manufacturers.

and, this is an open question - what manufacturer is going to let you enter the market like this? is campbell's going to let you do that? pepsi? dempsters? how long before they sue you for it?

and they should sue you and they should win - you're marking down their product at a loss to drive their sellers out of business, with the ultimate goal of becoming their new distributor by force. that's a damaging business practice. yes, that should be banned.

if i can convince you that the food industry is an oligopoly - and, like, read the fucking textbook. that shouldn't be a struggle. that should be an easy concession. - is it possible to convert it into a perfectly competitive market?

no.

it's an oligopoly for a reason.

and perfectly competitive markets are utopian. they're fairy tales. they don't exist in real life. approximations would be on very broad scales, like country a against country b. 

stores across the street from each other are nothing remotely resembling perfect competition. they sell the same shit that they buy from the same suppliers. they're both just factory outlets. 

the competition is between the companies that stock their food on the shelves, not between the stores.

if grocery prices are rising across the board - and they are - it's due to input costs such as:

- minimum wage increases in grocery stores

- labour costs at other places in the supply chains

- transportation and electricity costs

- fertilizer and pesticide costs

- tariffs on machinery

- etc

ugh.

price stability in an oligopoly works by ensuring that barriers to entering the market are high. this is a good thing, because it prevents companies from getting into price wars, which in the end results in bankruptcy, and price hikes when demand can no longer meet supply because the companies have driven each other out of business through competition, and are now all in debt, from trying to survive the price war. consumers may see prices temporarily fall, but it's at a loss to the business and it's not sustainable - it leads to steep price hikes in the long run, as the victors need to charge their way out of debt.

the end result of government policies that encourage competition is that the oligopolies turn into monopolies, businesses consolidate, an area with five stores loses two or three, and supply has difficulty meeting demand, leading to steep price hikes. it's carnage.

that is why oligopolies develop, because the companies calculate that is in nobody's interest - including the interests of consumers - for the businesses to compete. rational consumers don't want the chaos that comes from stores constantly fighting with each other, they want stable and predictable outcomes that they can budget around. they want the price to rarely if ever change at all. that's where price setting creates a superior and preferable outcome to competition.

if what these government bureaucrats are trying to do in interfering in the free market is create chaos and instability and drive stores out of business, thereby reducing consumer choice and driving prices up, it should keep it up. if it's trying to reduce prices, allowing new businesses to enter the market by undercutting the established ones will not do that, it will just create debt and bankruptcy and higher prices in the long run.

again, this is what you learn at a school when you enroll in an economics course. it's very different than what they talk about when conservative political clubs go get a beer, which is a bunch of nonsense that they go out of their way to educate out of you when you actually show up at the school and actually take the course.

again, this government is making economics mistakes you don't expect liberals to make. this is a giant leap backwards into retardo harperology land.

https://www.cbc.ca/news/business/competition-bureau-groceries-9.7360892