"Build, don't train" is poor advice for a prospective "AI Engineer" title. It should be "Know when to reach for a new model architecture, know when to reach for fine-tuning / LoRA training, know when to use an API." Only relying on an API will drastically reduce your product differentiation, to say nothing of the fact that any AI Engineer worth that title should know how to build and train models.
Fair point! I think my main idea was "prefer building with an API over training your own model" but that isn't as pithy.
The jury's still out on how much training and fine tuning are going to matter in the long run - my belief is that there are many great products that can exist without needing a new model architecture, or owning the model at all.
That advice makes sense if we're talking about 800B+ parameter models that require a gigantic investment of capital and time. For models that fit on a consumer GPU you're leaving chips on the table to not take advantage of training / fine-tuning. It's just too easy and powerful not to.
You absolutely lost me personally with "raises 2.5 million in capital"
History tells me to never trust anything this well funded for something likely to be so personal. Perhaps theres a business case and good luck, but I'll be avoiding.
Thanks for sharing original link — had not seen that project before. I’ll refrain from commenting on their funding, but fyi is mmm.page is wholly bootstrapped (which has left a healthy dent in my savings over the past three years haha).
So if you enjoy the project, or want to support some of the researchy aspects of it, please consider upgrading!
I have another article in the works about the importance of independent funding models and slow growth incentives w/r/t organic social networks, and projects that necessarily require slow exploration as a community (not, in contrast, “spaces designed by committee”).
Do you have any recommendations on libraries for parsing CAN with Rust (or are your tools OSS)? I'll be working on a project in a couple of months that'll need to pull data off CAN.
The article's thesis is that EA's moral realism stance (which finds it's apotheosis in utilitarian moral calculus) is ill-supported and highly questionable, especially when combined with theological anti-realism.
It's definitely not written to win over any EA supporter, that's for sure.
Is EA predicated on moral realism? That seems like a straw man. I don't see any contradiction between abandoning realism (i.e. admitting that there is no objective way to measure goodness or other moral qualities) and EA.
We probably just define EA differently. When I'm elected dictator, it'll be illegal to write an essay like this without supplying your working definition of the thing you're against.
In my view, EA generally means "taking utilitarian ethics ('the greatest good for the greatest number (of humans, or perhaps of other lifeforms)', with (premature) death as the worst outcome) literally and seriously in philanthropic work, usually to the exclusion of other philanthropic goals (aesthetic, religious, etc.)". Perhaps there is a form of EA that isn't at its core utilitarian, but then it's just reduced to technocratism and seems unlikely to inspire the almost religious devotion that EA current does.
Religious devotion? No offense, but that sounds like more straw. I think there's an awful lot of people who read about EA, visit some charity ranking websites, rethink their giving, and then don't go on to start a blog.
I don't think EA is utilitarian. Or at least, not in any way that requires objective moral realism. EA doesn't say, "moral goodness is quantifiable and I can prove on graph paper that one life saved from malaria is equivalent to 137 meals delivered to hungry people." It says, "I don't know what goodness is or how to measure it, but I need some way to decide where to donate, so..."
In other words, EA demands that you actually think about where your money goes and whether that's the best place for it, rather than just donating to whichever charity has the most emotionally moving imagery and music in its TV commercials. Once you do that, it naturally follows that almost any system you use will give the same answer: "Sheesh, I should probably donate less to the animal shelter." And I think 99% of the hate EA gets is rooted from the fact that people like donating to their local animal shelter, but can't really justify it.
EA was definitely founded by reasonably orthodox utilitarian philosophers, like Peter Singer, Peter Unger, Toby Ord, and William MacAskill.
Formal EA organizations and conferences also have a lot of utilitarians heavily involved.
So I think it's fair to say that EA is rooted in utilitarianism, although those same utilitarians might say, for utilitarian reasons, that it's a good thing when people think carefully -- or at all! -- about how to be more effective and have a higher impact, regardless of what metaethical views (if any) those people subscribe to.
I mean, fair, but does it follow that someone who dislikes utilitarianism or feels it's disprovable must consequently dislike EA? Because that seems to be the thesis of this essay.
I guess it comes down to this: when people argue with utilitarianism, in my experience it's specifically the part about treating strangers and friends equally, i.e. the idea that it might be better to save two drowning strangers than to save one drowning friend. I don't think EA relies on that or is based on it; you can have a value system that says it's better to save your drowning friend than to save ten or a hundred drowning strangers, and still find EA valuable.
If you dispense with realism, why couldn't you embrace the animal shelter?
Just bite the bullet. Say, "I've grown disenchanted with people, they suck. I'm going to dedicate myself to improve the lives of animals instead, who are mostly blameless victims of humanity, and are much cheaper to make happy than a human".
Utilitarianism doesn't really tell you how many dogs is a human worth, so you well can decide that you value dogs very highly, and that's all that's needed to make donating to an animal shelter very rational.
That's the thing, I think you can. If everyone who donates to animal shelters felt as confident about it as you described, I don't think anyone would have a problem with EA. Certainly, no one is complaining about the idea of, between two charities who do the same thing, giving to the one who does it more effectively.
But there's a lot of complaining about the idea of comparing two different kinds of altruism. And the whole point of EA, I think, is that the fact that it's uncomfortable leads people not to do it at all, and prodding them to think about it feels like an attack.
That assumes a tie-breaking morality that says objectively that animal shelters aren't a good enough cause. If that's your mindset, then no wonder people who donate to animal shelters don't like you talking about EA : - )
I think a more useful definition would be about instruments, not ends. It is up to each individual to define their utility function. If they want that utility function to be aesthetic or religious, that's up to them. EA as a philosophy doesn't determine what your utility function ought to be.
What it does say is: conditional on selecting a utility function, you should attempt to use the tools of rationality, statistics, etc to maximize it efficiently. The fact that particular groups of EAs have sufficiently similar utility functions to pool resources in determining how to maximize them does not mean that all EAs must agree on those utility functions.
If your utility function says the greatest good is converting non-believers to flat earthism, then EA says you ought to study how best to achieve that before devoting lots of resources blindly. It doesn't tell you that flat earthism is bad.
I think I've seen Mr. ACX say that this is wrong and the right way to do it is to be utilitarian, but just don't try very hard to calculate it ie don't be rational or maximize it.
This is probably correct; if you maximize anything hard enough you'll end up in a logic trap, turn the world into paperclips, and get eaten by a utility monster[0]. That's because rationality doesn't actually work in the real world[1]. Of course, "just don't try hard" isn't going to work for everyone.
I think if you do try to rationalize "utilitarianism but not calculating too hard", you end up at virtue ethics, but ironically I haven't thought about that hard enough to say.
[0] ie you can increase happiness temporarily by giving everyone drugs, they'll just die after
A reasonable short (so necessarily lossy) version of what I've ended up believing is:
- policy decisions should be made on a consequentialist basis (which can involve utiliarianism like calculations but is to me most about considering second and third order consequences, with attempts at calculation as merely one mechanism through which to do so)
- personal decisions should be made on a virtue ethics basis
I'm not going to claim this as a Right Answer but this model has proven to be the least wrong / most useful one I've encountered in practice, at least for me.
IMO that's a very reasonable approach with two caveats:
- Consequentialism has real issues with ethics in the very large to infinite scale (time, numbers of people, etc.). Most theories of ethics have issues in this limit as well, but Macaskill accepts the repugnant conclusion and advocates for "longtermism", which I think is undeveloped in anything more than a vague "we should think about the future" sense.
- It's perfectly valid for a human society to decide to have some non-consequentialist bright lines they won't cross.
I ... ok, so the late reply is because I know inside my head what I mean, but expressing it is hard (sorry about that, hopefully you'll see this anyway.)
Virtue ethics is ... about trying to reach for the Good. People of faith regularly have a similar outlook except perhaps without one of the 'o's.
Caring about people in general, trying to be trustworthy, trying to be honourable, trying to look after your friends.
I guess you could say it's characterised by trying to come away from a situation feeling like you've done the morally virtuous thing (I am very aware of the tautology there) rather than necessarily the utility maximising thing.
It's very fuzzy, but also very human, and I'm conscious that I've not given you a solid meta-ethical grounding for it ... but then again, I'm not convinced there -is- one, and yet it seems to work for me.
Sorry for the massive vagueness, but realistically it cashes out to "try to do the right thing", which is necessarily vague and dependent on your own perception of what is morally correct ... which is why it's a terrible way to make policy (IMO) but works well on an individual basis, at least for me.
> Virtue ethics is ... about trying to reach for the Good.
I guess the issue I have with virtue ethics in concept is that it's fuzzy enough that I'm not clear on how it sets itself apart against deontology and consequentialism. The way you put it, it seems deontology-lite.
> Caring about people in general, trying to be trustworthy, trying to be honourable, trying to look after your friends.
So it works out to sort of deontology, except less rigid maybe? Rather than a flat command of "tell the truth", you have a fuzzy idea of "try to be trustworthy"?
Do consequences matter? That is, what is important in the end, that you did the work of trying to look after a friend, even if in the end your effort made their lives worse, or whether your friends are happy?
I guess you could say it's characterised by trying to come away from a situation feeling like you've done the morally virtuous thing (I am very aware of the tautology there) rather than necessarily the utility maximising thing.
> I guess you could say it's characterised by trying to come away from a situation feeling like you've done the morally virtuous thing (I am very aware of the tautology there) rather than necessarily the utility maximising thing.
I think any utilitarian has to recognize that utility can't be predicted very far. Eg, the utility of a backup generator is highly dependent on whether you ever end up needing one, which highly depends on the local weather. So it makes no sense to obsess over it like a robot, you can't get there anyway.
The way I see it is that the split is about what matters in the end.
In consequentialist frameworks, it's the consequences. If everyone is happy, then there's no problem to be solved. On the other hand, if something horrible happened when somebody tried their best, they still did a bad thing in the end.
In deontology, it's the reverse. If you followed the rules, your conscience can be clear even if something bad did result. If you didn't, then you failed in your duties even if disaster was averted.
If I can ground why I find it useful at all, it would be in the intuition (and N=1 anecdata) that a heuristic of 'following the rules most of the time will, in fact, produce decent consequences overall.'
Caveats to that:
- I reject deontological absolutism as ridiculous and counterproductive.
- I can't really give you a solid idea of what my version of 'the rules' are and I certainly can't justify them beyond 'they feel right to me.'
With respect to the last part, I try to live by one of the tenets of Sartre's Existentialism and Humanism - "make the best choice you can with the information you have available, and if you did that, your conscience should be clear even if something bad results" with the caveat of "but you should probably still be -annoyed- that you got it wrong, and your conscience should only remain clear if you make an effort to understand how that happened and not repeat the error."
I very much have consequentialist -instincts-, but on a personal level I've effectively derived a form of virtue ethics from those instincts because I find it a more effective way to get an overall net good set of consequences than brute force calculation.
Fundamentally my choice to use of a form of virtue ethics for decision making is instrumental, not epistemic.
An example: as a general rule, lying is bad ... but keeping your mouth shut and/or omitting details is often useful, and I will happily lie to keep somebody's secret (e.g. avoiding outing somebody) both because of the desire to be trustworthy -and- because 'avoiding harm coming to your friends' is a virtue in and of itself.
That's all well and good, but I'm more interested in the practical example of how this particular person approaches it.
The Stanford Encyclopedia gives several forms for instance. It seems probable any given person will pick one form to subscribe to, and then have some particular interpretation of what that means to them. That's what I'm curious about.
Movie Thanos' plan is to kill half the universe to solve the resource problem. This seems obviously stupid to me, since the world population can double in about no time at all, and will create a huge incentive for growth. Imagine half of the housing in the world being empty. Great time to start a new family!
Comic book Thanos wants to kill half the universe to woo Death, as in the female deity representing death. No altruism in that as far as I can tell, just the world's most brutal romantic gesture.
EA adherents can use rigorous reasoning to reach different conclusions, just from having different values that they start with. Case in point: One person might want to maximize quality of current lives; another might want to maximize quality of future lives. Reasoning validly from there, they can reach wildly different conclusions on how to most effectively give.
> realism (i.e. admitting that there is no objective way to measure goodness or other moral qualities)
Moral realism is the position that mind-independent moral facts exist; it doesn't entail any particular belief about how or whether we can come to know them.
Who is this written for then? It seems to me that the author just builds up a few explicit strawmen and expects that the reader is already on board with the idea that they are obviously idotic.
It's not really against EA then, it's against all morality and ethics as such. Rightly title the article "Against trying to be a decent person" and it's obvious why it's a piece of junk.
The study only checks for myocarditis diagnosis codes 1-7 days after vaccination. That's a very short timeframe... We have multiple months worth of data at this point, I wonder what happens to the incidence rate at longer timelines.
I’m not a doctor but fairly well versed in inflammatory conditions for personal medical reasons. Myocarditis is primarily an acute condition. According to Google its onset is quick, and most people don’t suffer chronic forms. That’d imply that if myocarditis doesn’t happen in the first 1-7 days it seems very u likely it’d occur after that period. https://www.myocarditisfoundation.org/research-and-grants/fa...
I had viral pericarditis after a cold one time, but it appeared about 2 months after I had recovered.
That's pretty common with autoimmune conditions triggered by IgG autoimmunity response.
But we understand autoimmunity in response to individual exposure to antigens pretty well and while 1-7 days is probably too short to see a lot of responses, over 3 months is highly unlikely. There can be a long tail of certain individuals having autoimmune responses up until 2 years later, but in a population study if you capture nearly all the responses after 3 months. If you don't see anything 3 months later then there's no hidden long-term effects waiting to happen years later.
And there's nothing fundamentally unique about the mRNA vaccinations that would cause anything different. It is just a dose of mRNA that looks like any other mRNA payload from a virus which is wrapped in a lipid package that fuses with the lipids in your cell wall (and we understand the allergy to PEG that some people have against the lipid nanoparticle itself). All of the 200 years of understanding of vaccinations and hundreds of years of autoimmune conditions subsequent to viruses still apply to the mRNA vaccines. The "no we don't understand anything about these new things and the clock starts from zero" is just a fantastically ignorant argument based on no understanding of what the vaccines are built from and what vaccines and viruses are and how they interact with the immune system.
So both of you are kind of wrong. 7 days is too short, but 90 days is all you need.
Thanks, good to know about pericarditis. I’m not well versed on that or myocarditis and in particular the IgG or T-Cell response mechanism. Though the little I’ve read seems to indicate myocarditis is still primarily due to an acute immune response, but can re-occur or turn into a more autoimmune condition (chromic myocarditis or perhaps pericarditis).
That said I would normally give the benefit of doubt that the authors of a paper in the Lancet are more well versed on myocarditis and it’s behavior. It looks like they do adjust all their comparisons to odds ‘per 100,000 person-days’. If done properly that would adjust from an event basis to per-time basis or rate which could account for long tail reactions.
Using 90 days would risk including non-vaccine related myocarditis risks as well, which is perhaps why the lancet paper uses the time rate basis. If someone got a normal cold after a month and got pericarditis that'd skew the results as an example. Though I was disappointed to not see statistical analysis or comparison to normal occurrence rates of myocarditis without vaccine exposure, etc. though I didn’t thoroughly check their references, etc.
> The encounter methodology identified 14 distinct patients who met the confirmed or probable CDC case definition for acute myocarditis or pericarditis with an onset within 21 days of receipt of COVID-19 vaccination. When we extended the search for relevant diagnoses to 30 days since vaccination, we identified two additional patients (for a total of 16 patients) who met the case definition for acute myocarditis or pericarditis, but those patients had been misdiagnosed at the time of their original presentation. Three of these patients had an ICD-10-CM code of I51.4 “Myocarditis, Unspecified;” that code was omitted by the VSD algorithm (in the late fall of 2021). The VSD methodology identified 11 patients who met the CDC case definition for acute myocarditis or pericarditis. Seven (64%) of the 11 patients had initial care for myopericarditis outside of a KPNW facility and their diagnosis could not be ascertained by the VSD methodology until claims were submitted (median delay of 33 days; range of 12–195 days). Among those who received a second dose of vaccine (n = 146 785), we estimated a risk as 95.4 cases of myopericarditis per million second doses administered (95% CI, 52.1–160.0).
One of the author also did a podcast where she pointed out how the way some insurance claims (especially those from outside facility) works is that it takes a few weeks to receive the claim and it to get registered correctly. So they often get missed in short period windows.
When they took into account the correct numbers, they found rates of 537.1 per million for MALES aged 18-24.
Here in Ontario, our government reports rates of 200.2 per million in males aged 18-24. Note that Ontario stopped giving Moderna to under 30 year olds and also made the dosing interval to be 8 weeks but our rates of myocarditis continued to climb.
> most people don’t suffer chronic forms
This is misleading. There is no such thing as mild myocarditis. Inflammation of your heart muscle has to be taken seriously, even if you end up in the hospital for a couple days, that's not the end of inflammation because it leads to scarring and permanent damage. Heart is the only organ that's going to keep you alive the longest and any damage to the heart has to be taken seriously especially if you are balancing it with a condition that in itself is very low risk to youth.
Around 88% of the cases occur in males and up to 96% of the cases are hospitalized. Requires no exercise for 6 months and 6 month follow up Cardiac MRI still shows damage:
And since the COVID vaccine doesn't prevent infection, you may end up compounding the risk if you get myocarditis first from the vaccine and then again from infection.
The largest study on this topic has been of 23.1 million residents across four Scandinavian countries — Denmark, Finland, Norway, and Sweden. They found that:
> in young men 16 to 24 years of age, the risk for myocarditis after vaccination with either the Pfizer or Moderna vaccine is higher than the risk for myocarditis after COVID-19 infection.
> "We show that the risk of myocarditis after COVID infection is lower in younger people and higher in older people, but the opposite is true after COVID vaccination, where the risk of myocarditis is higher in younger people and lower in older people"
> "In the group at highest risk of myocarditis after COVID vaccination — young men aged 16 to 24 — the Pfizer vaccine shows a five times higher risk of myocarditis versus the unvaccinated cohort, while the Moderna vaccine shows a 15 times higher risk"
This finding from Scandinavian countries also seems to match the update to the Nature study which found the risk of myocarditis in under-40 males to be greater from the vaccine than from COVID. Discussed here few months ago:
I only have time to checkout out your first article. It seems plausible the KPN folks might have caught a tail-end effect of myocarditis in Covid vaccination. However the Kaiser numbers don’t appear that different or concerning to me as an educated lay person. They’re not an order of magnitude higher AFAICT without re-normalizing the units used in the TFA and the KPN paper. AFAICT from the error bars on table 2 in the KPN paper they have pretty wide error bars and CI. That makes sense given that even with on the order of a 100k+ vaccinations you only get 16 cases occurring.
TFA presents their data in a form seemingly normalized to 100,000 person-days not just events-per dose. This would potentially also capture much of the long tail effect. Though I’m not well versed in the odds ratio statics most common in these sort of medical studies.
Now that being said I am actually somewhat concerned of the lifetime cumulative risk of specifically Covid-19 vaccinations. Particularly if it becomes standard practice to get a new booster every 6 months for the rest of your life. Luckily that practice seems to be ebbing as Covid-19 has become endemic. TFA reports higher rates after the second dose, and excluded booster shots AFAICT. It’s possible booster doses could present the same or higher risk of myocarditis in particular if you’re targeting the exact same spike protein. Some quick back of the napkin calculations show a lifetime risk approaching the fatality risk of say omicron of you did get a dose every 6 months, and the risk was constant. Both of those conditions seem unlikely. For example susceptible people would likely be “weeded out” and the myocarditis risk would drop, and it’s doesn’t seem likely people will gets bi-annual boosters. Also the dosing of mRNA vaccines will likely get better and methods for treating rare events like myocarditis will improve.
> I am actually somewhat concerned of the lifetime cumulative risk of specifically Covid-19 vaccinations. Particularly if it becomes standard practice to get a new booster every 6 months for the rest of your life
I think it's valuable to have this public debate in general, and also regarding the quarantine measures. Quarantine measures have been criticized as something of a ratchet that doesn't have anything to do with disease, but rather with authorities controlling the population.
And I guess in some locales that criticism has strong merit. But in my locale, all the extraordinary laws enacted to mandate social distancing, quarantines and so on were rolled back this spring, triggering a big Covid wave that turned out to be (in line with health authority expectations) relatively harmless due to vaccinations and previous exposure. Some businesses struggled with their headcount for a few weeks, but within reason. And the laws weren't re-instated. So I recently caught covid along with seven others at a small party, where someone was coughing without covering their mouth all night and of course no one considered wearing a mask. But this is now all in line with official policy. Literally forbidden one year ago, on multiple counts, in principle punishable with jail but generally fines.
So I do think it's important to follow up on these arguments. It's always a cost-benefit judgement, and it's ultimately a democratic, political question. Not something that can be decided by a bureaucrat. I've always figured that everyone getting booster shots every six months for 30 years sounded almost parodically excessive, but certainly there have been a few voices demanding that.
> relatively harmless due to vaccinations and previous exposure
This year's dominant strain was significantly less lethal/debilitating for unvaccinated people as well.
The restrictions are still reasonable in hindsight, especially in places like Canada where hospital capacity was quickly overwhelmed in all waves leading up to and until this Spring.
Restrictions obviously can't create a world with "COVID zero" but it can slow down the infection rate so that hospitals can still treat other emergencies (although, again, in Canada it was barely enough). Kinda hard to run a hospital as usual if you have 30 people per day rolling in and they never leave.
This year's COVID waves are quite harmless also to unvaccinated people. It's because Omicron is less dangerous, not because of vaccination, but of course public health authorities will ignore this and claim victory. What else can they do? Public health is a dangerous concept by construction because the social costs of them admitting they were wrong are so high.
when surmising that the only reason someone hasn't "admitted" they were wrong (and assuming you are right), is fear of the social costs, it's always good to take a step back and see if you are actually the one unable to admit being wrong, for the same reason
What social costs? I'm a nobody on the internet. Although outside of HN I've posted things about COVID and the failures of public health, they had little impact because I'm not a part of that community. As a consequence:
1. I have no career in public health or academia to defend.
2. If I was wrong (I'm not) then it might hurt my standing a bit, but lots of people on the internet have opinions on random topics that other people think are wrong and they still manage to get jobs, create products, get customers etc.
3. Likewise, being right also doesn't help my career or life very much.
Contrast this to people like Ferguson or Fauci:
1. Their entire professional existence hinges on the perception that they are experts in disease.
2. If they were wrong (they are) then it'd not only hurt their standing a bit, but cause hundreds of millions or billions of people to hate them viscerally due to the staggering costs of what they insisted on. It would also destroy their self-esteem and result in them knowing that they'd done evil things, making them some of the most notorious people in history.
3. In contrast, if they were right, they .... get to keep their current jobs, salary and social status, more or less.
Frankly if I did work in public health it'd be very hard to imagine ever admitting to mistakes either. Public health operates through government force and COVID policies in particular involved forcing people to do things harmful to themselves at great cost to themselves, against their will. You have to believe you made the right calls if you're in that position, it's the corrupting influence of power. If you didn't you'd immediately quit in shame and never show your face in polite society again, so the only ones that survive in that world are the merciless ones who can never admit they are wrong. Which is exactly what we see!
this would be a good question to ask the experts you are accusing of dishonesty, so they have the opportunity to come up with a similarly defensive response based on equally plausible denials
take your example of Fauci for example - he's admitted mistakes, and has not lost the trust of most Americans like your premise claims (save for a minority who thinks mistakes, something all humans commit, are a sign of weakness).
so already we know the "social costs" for him that you're talking about as an example, are far overblown
indeed, most Americans understand the trade-offs that we made during the pandemic based on what we knew for sure at the time (not what internet people were claiming without trustworthy, peer-reviewed studies).
"he's admitted mistakes, and has not lost the trust of most Americans"
Has he? Which mistakes? I've never seen him admit any of the actual big mistakes like:
- Lockdowns didn't work (now the position of the German government's official enquiry).
- Masks don't work either.
- Deliberately lying about scientific topics to try and manipulate behaviour wasn't a good idea (he's admitted doing this to the New York Times, twice!)
- The vaccines were massively oversold.
Even Bill Gates has admitted the latter, but I don't recall Fauci ever doing so.
As for not having lost the trust of most Americans, maybe so, but that's just a testament to the shocking level of propaganda he benefits from. He's certainly lost the trust of anyone paying attention. The man is a self-confessed serial liar: he lied about masks, and he lied about herd immunity thresholds, both times making scientific claims about supposedly medical facts and then later changing his position, saying he was just trying to manipulate people's behavior. This is all on the record. Anyone who continues to believe this man about anything is certainly an idiot - fool me twice, right?
It sounds like you've already made up your mind on most of these things, and don't want to be convinced. For example, the U.S. never had an actually-enforced nationwide lockdown, and many conservative politicians and their supporters actively worked to make such a plan fail before it could even happen.
I'm impressed by the number of specific studies and writings you've gathered which happen support your conclusion. Tell me, what evidence and studies have you collected and evaluated which do not support it?
This comes across a bit disingenuous. Do you have evidence to the contrary? If so why not state what it is or that such evidence exists? I believe the Earth is spheroidal in shape and also haven’t collected evidence that supports it being flat. It’s not reasonable, in general, to look up the pros/cons of each belief. In the present case are you suggesting that people are cherry picking the studies to conform to a preconceived belief?
Yes, to your questions; however, do note that I am not the one publishing this unsatisfactory metastudy. As such the burden is on the publisher to produce something of value by looking up the pros and cons and comparing the two. Otherwise, anyone could just punch up google, print out the results for "flat earth" and claim to have done research which says you should believe it. Obviously, that's not the case.
This is not an academic forum and so the standards are a bit relaxed. Typically people post their beliefs along with a snippet of the reasons for said beliefs. Then others who disagree sometime respond by presenting contrary evidence. No one has a burden as such. We are not academicians beholden to standards of thoroughness.
My suggestion is that had you responded by stating that there are contrary studies and that those studies are better then your post would have been a lot clearer. At least it would have put the idea that the evidence presented was suspect into my mind. As it is it came across as unclear who has the stronger position objectively speaking.
Yes, of course, you're right. Unfortunately, Brandolini's Law often applies quite strongly. I simply do not wish to always subject myself to that and only wish to alert readers of the comment thread to think critically about the parent comment's contents. That is - the target of my comment is not the parent comment's author, but the reader of the parent comment. Generally, this is sufficient to meet my aims.
I understand that. I’m suggesting changing the wording. Your original post would have been a lot more helpful to people like me had it been of the form:
People should know that there are lot of studies and evidence to the contrary that is easy to find. The evidence presented above is flawed and you can look this up easily on your own.
My comment, as is, captures the tone I intended. Therefore, I will not edit it. Thank you for your kind suggestions. I will take them into account in future interactions where they are relevant.
What conclusion, besides the one you projected? He responded to a bunch of separate points the first of which actually addresses 'other studies.' Did you read it or have any thoughts on it?
I've read a number of these studies. My thoughts are that a number of them have confounders which are not well controlled for, and they often have limited statistical power, representation issues, or other technical faults which impact their expected validity. Further, in most cases, reproduction of the study is generally not possible. In any case, individual studies simply listed supporting a particular conclusion particularly without a listing of similar studies which present opposite conclusions - and a direct comparison therein - does not a metastudy make. What's attempted here is a metastudy and, as a metastudy, it's quite lacking.
It's an extremely important question to ask and especially when someone dumps a wall of text and a ton of studies on you that all support one viewpoint:
* Have you also evaluated studies that don't support this?
* If not, why not? Are you only looking for studies that support what you want to prove?
* If you have also looked at studies showing negative results, why have you decided not to include them here?
myocarditis isn't generally chronic, particularly in kids. and that study concludes that the risk of the vaccine in 12-19 year olds outweighs the risk of the virus if you bother to read it.
No, if you read the article it's from 12 up the benefits outweigh the risks, quoth:
The ACIP report projected that mRNA vaccination in 12-17-year-old males would result in 215 fewer hospitalizations and 71 fewer intensive care unit stays. Benefits of the vaccine outweighed the risk of myocarditis from vaccination in all age groups, 12 years-old and up. Our results suggest that the risk of myocarditis from COVID-19 infection itself exceeds the known risk from vaccination by a considerable margin. In light of more infectious variants, the new school year nearing and many colleges now requiring COVID-19 vaccination (either for all students or just those living on campus), these results are especially timely. Whether considering all the risks and benefits of COVID-19 vaccination or just myocarditis, vaccination appears to be the safer choice for 12-19-year-old males and females.
Certainly we can say our ML models are becoming more general in the sense of being able to cross-correlate between multiple domains. This is quite a different story than "becoming a general intelligence." Intelligence is a property of a being with will. These models, and machines in general, do not posses will. It is we who define their form, their dataset, their loss function, etc. There is no self-generation that marks an intelligent being because there is no self there at all.
It is only the case that ML expands our own abilities, augments our own intelligence.
Assumption of will is unfounded, scientifically speaking. Your entire argument is philosophical, not scientific. The subjective experience of free will is in no way unrefutable proof that will is required for intelligence.
Since a working (in the sense of 'working title') ontology and epistemology are _required_ for science (read "natural philosophy") is this argument not arguing that "the argument for quarks is unfounded, biologically speaking"? That said, I _believe_ that both Aristotle and St. Thomas agree with you that will and intellect are not necessarily connected, so you could have an intellectual power with no freedom to choose.
This points out something very related that I think about a lot - can you prove to me that you do any of those things? Can I prove to you that I do any of those things? That either of us have a will? When would you be able to believe a machine could have these things?
In Computing the Mind by Shimon Edelman is a concept that I've come to agree with, which is at some point you need to take a leap of faith in matters such as consciousness, and I would say it extends to will as well (to me what you've described are facets of human consciousness). We take this leap of faith every time we interact with another human; we don't need them to prove they're conscious or beings with a will of their own, we just accept that they possess these things without a thought. If machines gain some form of sentience comparable to that of a human, we'll likely have to take that leap of faith ourselves.
That said, to claim that will is necessary for intelligence is a very human-centered point of view. Unless the goal is specifically to emulate human intelligence/consciousness (which is a goal for some but not all), "true" machine intelligence may not look anything like ours, and I don't think that would necessarily be a bad thing.
Not just consciousness- all of science requires a leap of faith- the idea that human brains can comprehend general universal causality. There is no scientific refutation for Descartes' Great Deceiver- it's taken as a given that humans could eventually overcome any https://en.wikipedia.org/wiki/Evil_demon through their use of senses and rationality on their own.
I have long worked on the assumption that we can create intelligences that no human could deny have subjective agency, while not being able to verify that. I did some preliminary experiments on idle cycles on Google's internal TPU networks (IE, large-scale brain sims using tensorflow and message passing on ~tens of pods simultaneously) that generated interesting results but I can't discuss them until my NDA expires in another 9 years.
Dunno. For all I know, DeepMind will publish a paper titled "Evidence of Emergent General Intelligence in Deep and Wide Perceptron Hypertoroidal Neural Message Passing Networks of Tensor Processors Trained over the Youtube Corpus" and get all the credit. :)
Which brings up a related tangent. How is it Deepmind has a veto over what you publish? I understand keeping proprietary knowledge and implementation details secret (though the industry trend is in the other direction), but forbidding publication of your research seems excessive.
Appeals to humanity do not convince me of anything. I do all those things (well, I dance terribly) but again, those are not indications of will, and it's entirely unclear what magical bit in our bodies is doing that, when computers cannot.
Even if you don't want to have anything with such a science, such a science will move on without you.
"A version of an oft-told ancient Greek story concerns a contest between two renowned painters. Zeuxis (born around 464 BC) produced a still life painting so convincing that birds flew down to peck at the painted grapes. A rival, Parrhasius, asked Zeuxis to judge one of his paintings that was behind a pair of tattered curtains in his study. Parrhasius asked Zeuxis to pull back the curtains, but when Zeuxis tried, he could not, as the curtains were included in Parrhasius's painting—making Parrhasius the winner."
Why would an AGI be unable to do these things? Sure, if you believe in a transcendental soul (mind/body dualism) then you can argue that it can't because Divinity has simply not endowed it with such, and that claim can neither be proven nor disproven. But it's an extra assumption that gets you nothing.
Note that I personally believe we are more than a century away from an AGI, and think the current models are fundamentally limited in several ways. But I can't imagine what makes you think there can't be a Ghost in the Machine.
I've been publishing writing far outside consensus progressive attitudes since the pandemic began with zero issue so far. Don't make your workplace your audience or invite controversy around your workplace and you're very unlikely to have a problem.
Harm is done to the ecosystem when we concentrate cows in one place and feed them chemically fortified slurry. When cows are allowed to graze pastures, they instead have a positive effect on the local ecosystem.
As such, I am extremely skeptical that the right way forward is to reduce and concentrate the essence of a cow into bioreactors and feed them chemically fortified slurry.
I'm interested in those answers, too. I imagine it's much less waste because you can feed the cells just what they need to grow, instead of feeding the bacteria in the guts of an animal.
You ignored the part of the GP comment where they agreed with you. The point is that properly managed pasture increases fertility naturally. We need animals to fertilize, or every plant must be grown with synthetic products.
What purpose is served by ignoring any example of dehumanization?
What purpose is served by even attempting to downplay the "jury-less execution" aspect of armed drones, and pretending that it matters at all that these drones are not armed yet?
A few different purposes are served but no valid ones.
