59 karma · joined January 30, 2022
Could you give me a link to what you're referring to and what you have seen in the past with this specific problem?
I think I have it solved but would like to know more as words can have multiple definitions.
Most importantly, anyone can host this completely for free either on a local machine or a VPS.
The setup process includes a command line interface which allows you to enter a few options and it sets everything up for you...assuming you have a server such as Digital Ocean and a domain you can use.
I do this in my free time with some others who are of course listed on the Github repo.
We will be moving to a services based architecture at some point soon, but for this release it's basically a, "pretty good monolith."
We have been through a few different release cycles since the last time I posted. I added some GIFs to the readme to clearly show how easy it is to get it going. I would love it if anyone wants to give it a try. We included automatic SSL establishment via Traefik.
Thanks in advance if you choose to give feedback.
* Rome was incredibly paternalistic, so pointing out the failures of others without appropriate aristocratic station might not have been quite the exercise in pure engineering that we might consider engineering to be today, even though Rome obviously made great engineering accomplishments and one might think that celebrated Roman architects could not be writing anything other than in a strictly professional manner. I think that when we read Vitruvius through a modern lens, we might think, "Oh this dude was a real professional, he makes good points," when in reality I think, at least in the example I point out in my blog post, I think he might have been using an example of failure to say, "See? Only royals, aristocrats and rich people should even try to innovate."
Granted, there wasn't liability insurance for architectural or engineering failures as far as I am aware so this belief may have been prudent in terms of saving lives, but I think there was probably a primarily social-religious reasoning for it.
The idea of the rich, hereditary classes, the father of the household, the ones in charge being the only ones, "allowed," to innovate was deeply held in Ancient Rome, much like in Confucianism and other agricultural societies that have formed through human history. "Don't rock the boat."
I don't think it was until the 1500's or so that humanity started to collectively emerge out of that type of mentality as pre-industrialization and the scientific revolution began. Eventually the aristocrats (in Europe) saw non-royals being so successful at being industrialists, they wanted in on the game and removed laws that prevented them from engaging in capitalism.
The author writes:
> Its failure became a cautionary tale of an unrealistic project that sought to go beyond what nature allowed. Five decades later, the Roman senator and historian Tacitus claimed that evidence of this ‘futile ambition’ could still be seen in the rock faces near Avernus. Interestingly, Tacitus did not blame the vainglorious and weak-willed emperor alone, but also the architects who conceived the project and were judged to have ‘frittered away the resources of a Caesar’.
I read that as the author saying that Tacitus was angling the engineering failure analysis in a political way, not a purely engineering post-failure-analysis way. E.g., the project might have even been a success but then Tacitus just lied about it, we don't know.
Though I'm guessing that was a pretty expensive domain name. So in lieu of changing it, on this issue I can be bought off.
But I will accept a bribe in the form of a medium sized T-Shirt to not make fun of you further. The T-Shirt must include an anthropomorphized gaussian curve in a math meme format and the curve-person must be clearly labeled as being name, "Kurt Osis."
Any Kurtosis logo must be either non-existent or super non-prominent so that I can at least pretend that it's a meme and then maybe tail into where it came from. A huge emblazoned Kurtosis logo on the back or front will be unacceptable (although Kurt Osis could have that logo on his or her or their shirt).
* Devops folks don't seem to tend to like math and often got there by practicing, "computers, IT, having fun hooking things together and getting them running."
* Data science folks don't tend to like devops and prefer to bash around on a jupyter notebook that's already given to them and then maybe extract that python and see if it runs, but they tend to come from more of a science background and got into python as a hobby or incidentally. They do not like bashing around and getting things running.
So now this company is combining a math term that has a specific meaning to an ML ops space, which is going to cause confusion.
Different sets of data can have different kurtosis measures. Sets can be platykurtic (flat gaussian curve or high kurtosis) or leptokurtic (tall gaussian curve, low kurtosis).
Now this company is coming in and telling a bunch of devops people, "Kurtosis means helm but automatically migrate data too." So they are applying the idea of, "leptokurtic deployments," presumably with the metric being, variation between the code and data parameters on those servers. Data science people who are told about it from devops people are going to initially hear, "somehow dealing with cleaning the data, like an ETL pipeline, perhaps an Airflow with data cleaning tools built in or something."
It's very confusing and not helpful to customers, I hate it. There are going to be meetings where ML/Devops people are very confused.
Naming is hard though -- but I wish they would have gone with something like, "platypus" and just have a cute little platypus baby as the logo and say, "yeah we liked the word platykurtic because we like making things regular and platykurtic sounds like platypus."
--> Have me command some Fatwas, Responsums or Dictums for you to issue
--> Interpret Religious Text For You
--> Recipe Ideas
I became more skeptical of this piece of propaganda when I saw the top comment from prominent COVID origins conspiracy theorist and statistics misuser on Substack (who evidently uses the pseudonym Yuri Bezmenov?) on the post, as well as likes from Academic Capture extroidinaire Richard Hanania, of the Salem Institute at UT-Austin (a business school whose name is emblazoned by a now dead billionaire to whom I as a Minnesota citizen, personally pay taxes to his estate to every time I buy food, in the form of subsidies to his Vikings stadium).
Basically I'm sitting here being lectured by people about Aristocracies while I, as a nobody schlub, pay taxes to a private party they are friends with every time I get groceries, and meanwhile see schools nearby me go on strike, there is no free daycare in Minneapolis while there is in other cities nearby because, "it's not fiscally responsible?" Ok, well I'm not an "economist," at an, "institute," so what do I know?
Anyway, I was very confused after reading this by the claim that Jefferson would not have considered himself part of the Artificial Aristocracy so I had to read about his family history. From Monticello.org:
https://www.monticello.org/research-education/thomas-jeffers...
> The origins of Jefferson's ancestors might be uncertain, but there can be no doubt that within a couple of generations the family had risen from the humble rank of "middling planter" to the county elite, and within another to the very pinnacle of society. Their spectacular rise in fortune was the result of hard work, advantageous marriage, and sheer good luck
So what I read this is, is his family were not part of the English Aristocracy, (like Washington) but go in early on Virginia Real Estate in the 1600s, which allowed Jefferson to become part of the Colonial Aristocracy (though not the formal English Aristocracy). So Jefferson is saying, "no, that Aristocracy is not cool, bro, our Aristocracy is the cool one."
It also seems that some of Jefferson's ancestors were very anti-monarchy during the English Civil War, and lost a lot of fortune upon the restoration of the monarchy, so it's super understandable why Jefferson would take this position.
The thing is, I believe Jefferson may have become a dinosaur in his own time, writing letters to Adams to re-assure himself, but actually just discussing the types of ideas that would have appealed to some of the characters straight out of Barry Lyndon. E.g., he probably was witnessing Aristocrats getting their heads cut off in France and going, "no we're not those guys, we're a natural Aristocracy, it's totally different man!"
Jefferson was a huge Ancient Rome nerd because he was interested in creating a society that was durable, where other societies had failed an while I believe that some of his ideas that lead to a sort of weighting of rural areas does provide stability, because cities always end up being more powerful than rural areas, I'm pretty sure that the 1800's and early 1900's was a round rejection of Aristocracy, but of course there's no way he could have seen that. Venture Capitalists, pretend Professors from UT-Austin business school who are hired help to form an intellectual framework behind Rick Scott's agendas, they know better than this, and they should do better. This is basically a circle jerk with almost zero self-awareness about how Universally hated Aristocracy of any kind is, it's like how John Bolton has been talking about running for President in 2024, there is absolutely zero self awareness about how hated and loathed a particular idea is.
Edit: Oh yeah, and also, you can't mention that Jefferson owned slaves and so maybe not every idea that came off of the top of his head should be celebrated because evidently just acknowledging that fact might be considered, "woke" according to Hanania, most likely.
Wrote more about that here: https://patdel.substack.com/p/it-was-good-talking-to-you-as-...
Now, since we all are adults here and we all understand that the media gets manipulated, what are some tools we can use to understand the real state of affairs?
At least in the United States, herd immunity has been largely achieved for the time being, in conjunction with treatments and lower case levels (regardless of testing), rendering COVID less of an actual threat to a wide audience of news consumers, which is actually an OK thing.
Healthcare systems are overloaded, in part due to an aging population, in part due to mass quitting during COVID, in part due to an obscenely unhealthy and obese population, and a population that is getting older on average. People who don't know about this yet will understand when their time comes to go to the Emergency Room and they get a huge massive kick in the pants having to wait for hours and hours, or die. So that's a separate issue in it of itself beyond COVID, it's just that if COVID comes back, or any other disease comes back, healthcare availability becomes worse.
So basically, if a thing is not objectively dangerous to a wide audience, it doesn't make as much sense to report on it, because it's less relevant. You can set up an alert condition, "Red," but if that alert condition is constantly, "Red," forever, then it becomes less actionable, particularly if the result of the underlying problem behind the alert is of lower consequence.
None of this is meant to say that resources or attention should not minimize the future threats of COVID or related disasters. None of this is to say that time and attention should not be spent on biological hazards, such as the chance of another pandemic springing up from zoological sources, or perhaps new variants of COVID mutating and becoming a problem again. Nor is this meant to minimize the very real problem that players with consolidated financial capital, (read: billionaires and corporations) manipulate the media, and that we should not read media blindly and unreflectively.
There's also of course the important story of healthcare effects toward vulnerable populations, and minority populations. If that's a separate story, it merits attention, but I definitely don't think raising a Red alert for all of society is a good way to solve that problem, as people just become skeptical of alerts.
But at the end of the day, if a problem of global climate change has once again become a time-weighted larger existential threat than a particular biohazard, then it makes sense to start reporting on it again, to a with a higher degree of importance or alarm. There isn't really a reason to be, "jealous," that an article on global climate change makes the round, while your chosen pet issue did not, because if your chosen pet issue really is a big problem, people are going to eventually get interested in it, almost regardless of how much money is spent covering it up. We live in a heavily manipulated environment, but it's not North Korea (yet).
(edited for readability)
The key here is that it's speculative.
That being said, a lot going on in AI, it's wild out there, difficult to know what's going on. I tried to encapsulate some of the ramifications of all of the Open Source stuff going on with this article here: https://patdel.substack.com/p/viva-la-revolucion-ofopen-sour...
What am I able to play with? Why does this say on the first page that it's an essay? Where is the link to be able to download and play with it? Am I missing something?
Marqt, much like prediction markets, is not seeking truth, it's creating a map of a showcase of whoever bought into a particular market's emotions.
It sounds like you're motivated by creating a repo of human feedback reinforcement data which could then be used to further train or fine-tune LLM's, perhaps even to sell, is that correct?
> important for society to accurately define truth in an age when information travels (and changes) faster than ever, and when LLMs so confidently lie and have been heavily prompted to avoid controversial topics.
Yeah, I don't think what you're building is getting at the goal of, "truth," it's getting at the goal of building something sellable per my question above. To the extent that someone will be willing to buy that human feedback database is a function of what they may be trying to accomplish.
Much like in the world of so-called prediction markets, I see several of your markets as being heavily biased toward speculation and feelings, as opposed to a pragmatic approach to predicting a future outcome. What I mean by that is, you could create a really excellent scientific model and way of monitoring under what conditions a dam will break, or you could poll local people around a dam whether they feel like it will break, and you may get wildly different means and variances from the two methods. In all likelihood, there is no extra new synthesized knowledge about empirically what will physically happen with the dam that a group of average people off the street will have about that dam, in the context of a society with functioning infrastructure and tens of thousands of working dams. What you will have is a database on people's emotions about the dam, or perhaps dams in general, or the company behind the dam, etc.
So what are you really doing to, "make the world a better place?" You're not, you're making it worse by selling a vision of, "truth," that is actually muddling certain applications of practical engineering or scientific predictions further.
I think if you instead tweaked your pitch and just honestly stated that it's a repo of groups of people's feelings on various topics (and who those groups are, to the extent that you can figure that out), and not a way to, "beat LLM's with real truth," then it would be a lot more responsible.
On the other hand, you could just be a troll trying to actively muddle the truth further, in which case you can just ignore my advice, that's fine too.
> Inside the marqt maker is a fine-tuned BERT model that roughly ensures the statement is sensical, grammatical, and not a question.
Can BERT conduct logical reasoning? So further to the above, it sounds like you're using an LLM to clean human feedback data, if that's the real crux of your plan, to make reinforcement data cheaper, and everything else is a sort of 4D chess, bravo, cool project!
What does Iowa, Wisconsin, Minnesota, Michigan have? They need to be really attractive places to live because they aren't inherently (though some parts are, they are really far away from population centers). People don't go on, "vacation to the Midwest," they go to other, better landscapes. I'm sure there are outliers and I know there are outlier locations, I'm speaking in generalities here. I have driven over a significant portion of the Midwest, it's few and far between. You can't arbitrarily make beautiful mountains pop up in Iowa, and it's cold. At least if you're in Arkansas, it's got the Ozarks, warm weather pretty much year-round.
So really the only thing we midwestern states can do is look deeply inward and improve ourselves, that's really the only thing, other massive or more interesting States can rest on their laurels for a bit because people are inherently going to want to move there. Can you tell I'm not running to be the Governor of Minnesota any time soon? ;-)
1. We rest on our laurels a lot and look down on the Southern States for having been, "racist," which really is a red herring for our own racism. "Look at what happened down south in the 1960s, thank goodness Minnesota was not like that." "We're never going to give back the 28th Virginia battle flag from the Civil War, Minnesota will be free forever!"
Yet, the Twin Cities has perhaps the largest education gap in the entire United States between black and white Americans. There are any number of explanations for why this happened, perhaps including redlining, or failed well-intentioned policies, etc., but it's a fact. I'm not defending actual racism in southern states in any way, but clearly we have something wrong here.
2. We're extremely patriotic, perhaps even more so than Texas, there are Minnesota iconography everywhere, extreme pride for all things Minnesotan, lake country, living through the cold, Sven and Ole jokes, etc. Yet there were literally almost approaching Nazi-level genocidal parts of our history which are reflected in our flags, our naming conventions, where Native Americans live, and a huge majority of people refuse to acknowledge this because it's not nice to be controversial and talk out of turn or, "be distracting from real issues." You can go to Berlin and see huge museums acknowledging the holocaust, the horrors of the past are widely acknowledged, whereas here in Minnesota I was never taught about what we did in our history, it was highly glossed over in primary school, I had to learn about the shenanigans a lot of the early Minnesotan leaders whose names adorn our streets and counties pulled through Wikipedia.
3. Overall, the disgraced Radio Personality, Garrison Keillor hit the nail on the head when he said, "All the Women are Strong, All the Men are Good Looking, and All the Children are Above Average."
As much of a creep Keillor was, that really sums up a large part the mentality very well. There's no need to improve because we're already so great, and if we aren't that great in that specific way yet, we will surely get there, so don't push on it too hard. We basically think of ourselves as a, "Scandinavian-level-democracy country," for some reason, yet, please pay no attention to any number of typical American problems that we have just like any other state.
This is not to say Minnesota doesn't do well, and that I don't like Minnesota, we do rank well in relation to other states on a lot of metrics---it's more the mentality that I'm ranting about which I think prevents us from doing even better, which I think the author is getting at in a more broader geographic sense.
Would be delighted to hear any feedback or shared info from anyone who has dealt with this problems previously. Thanks for reading!
* Monsanto was the primary U.S. manufacturer of PCBs until they were banned in the U.S. in 1977, having manufactured 99% of PCBs [1].
* There are ongoing lawsuits in the United States against Monsanto for the manufacture of PCBs. [2]
* David Carpenter has a huge amount of studies showing how bad PCBs are. [3]
* While I can't see any evidence of ongoing funding of Bayer to the University of New York / Albany, Bayer does fund scientific research, fellowships, etc. in Medical and Agricultural Science [4], which may be a sufficient explanation to build a plausible narrative that someone or a few number of people at that University could have potentially been afraid of losing some grant. I don't have direct evidence supporting this, but there is reasonable suspicion.
All of the above coupled with the fact that, at least according to Dr. Carpenter, that the University concluded that his work in testifying against this massive corporation was not in fact problematic. [5] suggests that instead the University at the very least was extremely careless in even temporarily disciplining and muzzling this guy. They could have allowed him to continue working while doing an investigation.
By the way, you characterize Dr. Carpenter's serious work as a, "side hustle," for some reason - I'm assuming you were just trying to be cheeky and funny, which is good and hilarious - we all need to laugh. However if you're trying to attack him by saying, "side hustle," - I would encourage you to read up more on this story and all of the researchers out there in history who have tirelessly worked to protect you and I from chemicals. For example, lead used to be normal in gasoline - we all know a-priori that lead in gasoline is horrible now, but we may not all be familiar with PCBs because it hasn't been covered in the news as much or whatever. Just pause and think a bit about this stuff because it could really seriously affect your health some day, cause you cancer, cause your family members cancer - stuff you would really care a lot about presumably.
[1] https://en.wikipedia.org/wiki/Monsanto#:~:text=Until%20it%20.... [2] https://oag.dc.gov/release/ag-racine-sues-monsanto-knowingly... [3] https://scholar.google.com/scholar?hl=en&as_sdt=0%2C24&q=dav... [4] https://www.bayer.com/en/societal-engagement/science-educati... [5] https://www.supportdavidcarpenter.org/_files/ugd/2cea04_bb14...
My contention, which I covered in this video below here, is that due to the underlying statistical sampling problems inherent in RLHF transformers, LLM's perform poorly in edge cases, which, depending on the application or language, the margin of that edge can be super wide.
Here's a video I created about it: https://www.youtube.com/watch?v=GMmIol4mnLo
I didn't cover this yet but there are these things called, "scaling laws," which basically state the amount of raw text needed for a LLM with of a particular size of parameters. So my current mental model is that these, "laws," are really economic rules of thumb, like Moore's law is actually Moore's Rule of Thumb, and there is a huge expense in sampling clean data, hence the need for RLHF.
More about RLHF if not familiar with that term yet: https://huggingface.co/blog/rlhf
[1] https://www.noaa.gov/education/resource-collections/ocean-co...
Sorry in advance for being long-winded here, I have been thinking about this a lot:
I have another channel where I sporadically posted life videos over time, one of which has over 250,000 views.
What I notice from transitioning over to actually attempting to create quality opinion pieces / reporting is that it takes a massive amount of thinking/resources. So if you have little to no subscribers, if you want a 500 view video, you probably have to put 20-40 hours in, at least for technical content, and it needs to be somewhat topical.
If you listen to a lot of the really high-end YouTubers on podcasts, they are insistent that the algorithm really doesn't matter as much as the video quality.
Starting out, YouTube is very conservative about sharing a video internally on its own platform, because they don't know whether your video sucks. You can get them to accelerate that sharing process by using other networks, Twitter, message to friends, etc., but ultimately if you don't show a super high average watch time, your individual video will not blow up.
So that's the difficulty as an individual creator, particularly just starting out - you have a huge hill to climb. I notice smaller individual creators, in the 10k or so subscriber range, hack this by just creating new content constantly on a weekly basis. I struggle with this because I believe a lot of that content is largely garbage, and I know that if I were to do that, it would likely also be absolute garbage.
Now contrast this to news media outlets, who have been doing this for decades - they know how to pump out super high quality stuff on a daily or nightly basis, they have paid teams, studio set ups, editors, the capability to buy footage, etc.
So basically, "the algorithm doesn't matter bro, it's about quality content," is true, but at the same time, the algorithm really does matter, because YouTube constantly needs to pump out new content, and the best new content, or consistent new content is from professional studios, because for them to put in 20-40 hours of work on a video on a daily basis is already their bread and butter.
Does that make sense? Or am I off base here?
Anyway I think the only way to combat it is the same as with any other social media site - you have to actively curate your subscriptions to the content you want. On YouTube this often means being super aggressive and saying, "don't show me content fro this channel," not just disliking a video. I believe YouTube is very aggressive about showing you content that your frontal lobe might not want to see, but it knows your amygdala really wants to see, YouTube doesn't really listen to your recommendation input very closely.