king07828

Should a proper large language model be able to generate arguments for and against any side of any debate?

Create a monopoly tax that incentivises large companies to "right size". E.g., if the company is larger than X and has captured a market (like google with search) then tax all revenue above a threshold percentage of the market. For example, the market is $100B and the company makes $90B revenue, then tax away everything above $40B in revenue. This may take away the incentive of a large company to take too much market share and may encourage them to enable competitors. The parameters (company_size, marketshare_percentage, tax rate, etc.) may be adjusted to maximize performance

Rights can already taken away from people who are declared legally incompetent [1]. Similar procedures could be used to give rights to monkeys that can prove mental competence (e.g., are able to meaningfully communicate and show an understanding of money and property).

[1] https://www.law.cornell.edu/wex/incompetence#content

A critique of most modern billionaires and millionaires is that they settle for ivory tower philanthropy throwing money around so they can chat with rich experts to figure out which other rich experts to give more money to. While good does come from this, it doesn't seem to make much progress in solving the underlying problems. For instance, how long has the Gates Foundation been involved with vaccines and pharmaceuticals and does it look like solutions are in within striking distance for the underlying issues?

As an example, a millionaire (who recently became a billionaire) decided on two areas where mankind needed dramatic improvement. Instead of ivory tower philanthropy, he started two different for-profit companies, one for each set of issues. Both companies have stuck around for more than 10 years and the issues they are tackling appear to have solutions within striking distance. A tangible difference has been made. If you didn't guess, the two companies are Tesla and SpaceX.

In contrast, it seems like the Gates Foundation is treading water with the issues it’s tackling. I could be wrong, but that is my perception.

As another example, an entrepreneur back in the day decided every desk needed a computer on it and used his for-profit company to make that goal a reality. That entrepreneur, of course, was Bill Gates himself.

Very few people have the R&D and engineering expertise, economic sense, and the money to lead a company to achieve world changing endeavors. Gates did it once. Musk has done it more than once.

Gates could start a pharmaceutical company, run it himself, learn the ins and outs, and reshape the industry along the way (if he could successfully learn the ins and outs). However, he does not appear to be willing to accept the risks involved in starting his own a pharmaceutical company and the problems of the pharmaceuticals industry remain unchanged.

One can argue that the conduct is detrimental to the victim's interests and that the conduct constitutes common law fraud [1] because (1) the saboteur tells the victim "I like you", (2) the saboteur knows this is false or reckless, (3) the saboteur intends for the victim to leave the victim's current relationship, and (4) the victim left his or her current relationship.

I am not your lawyer. This is not legal advice. If your partner is actively sabotaging your relationship, it may not be a good idea to stay with that person.

[1] in Texas: In order to prove fraud, a plaintiff must show that (1) the defendant made a material representation that was false, (2) the defendant knew the representation was false or made it recklessly as a positive assertion without any knowledge of its truth, (3) the defendant intended to induce the plaintiff to act upon the representation, and (4) the plaintiff actually and justifiably relied on the representation, which caused the injury. Ernst & Young, L.L.P. v. Pac. Mut. Life Ins. Co., 51 S.W.3d 573, 577 (Tex. 2001).

I'd like to see the LSTM brain in the OpenAI 5 model [1] replaced by GPT3 to see if there is any improvement.

[1] https://neuro.cs.ut.ee/the-use-of-embeddings-in-openai-five/

This is awesome. What a concept, identifying the variables in the equations and providing diagrams to show the geometric meaning of the equations. Really wish wikipedia (or some sort of companion website) would similarly present an explanation of the practical meaning of variables and operators in the heavy math pages so that people with an understanding of algebra could understand the formulaic descriptions in addition to the qualitative descriptions after landing on a random page with heavy math.

> Executive orders that deprive natural rights are unconstitutional and thus illegal.

This may be a reference to the unalienable rights to life, liberty, and the pursuit of happiness [1] which may be protected by the 10th amendment to the US constitution [3] (reserving rights ... to the people) and the due process clauses [4] from the 5th and 14th amendments to the US constitution.

> Which right gives you the ability to operate your business during a pandemic?

Since "No person shall ... be deprived of life, liberty, or property, without due process of law" [4], the proper questions may instead be: What law gives government the power to close down businesses during a pandemic and was due process followed in the passage and execution of that law?

Arguably, the shelter in place orders may or may not be constitutional. Practically speaking, telling people to stay home for a few weeks using the threat of the shelter in place orders most likely helped the medical community get a grip on the situation, regardless of the constitutionality of those orders, which may explain the lack of resistance in passing those orders. However, at what point is "the cure ... worse than the disease" [5], i.e., at what point will the overall net net net effect of maintaining the shelter in place orders be worse than rescinding those orders?

[1] https://en.wikipedia.org/wiki/Life,_Liberty_and_the_pursuit_... (citing to [2])

[2] https://en.wikipedia.org/wiki/Natural_rights_and_legal_right...

[3] https://en.wikipedia.org/wiki/Tenth_Amendment_to_the_United_...

[4] https://en.wikipedia.org/wiki/Due_Process_Clause

[5] https://en.wiktionary.org/wiki/the_cure_is_worse_than_the_di...

Elon Musk giving shade to space based power "You'd have to convert photon to electron to photon back to electron. What's the conversion rate?"

[1] https://www.popularmechanics.com/technology/a8101/elon-musk-...

One person's "abuse of power" is another's judicial activism [1].

MGM v. Grokster [2] effectively held that distribution for the purpose of encouraging illegal activity can create liability for that illegal activity. See MGM v. Grokster at 919.

Applied to your gun shop example, do gun shops distribute guns for the purpose of encouraging illegal activity? Proponents of MGM v. Grokster would say no and that this may be the very distinction that the court was trying to highlight. I.e., distribution for the purpose of illegal activity versus distribution for legal activity.

From this perspective, it seems Popcorn Time may not have learned from MGM v. Grokster. Specifically, highlighting and promoting the illegal uses of a product may incur liability for those illegal uses. It may not be a silver bullet, but highlighting only the legal uses of a product may reduce the chance of MGM v. Grokster from being used to create liability for a product that has both legal and illegal uses.

[1] https://en.wikipedia.org/wiki/Judicial_activism

[2] https://scholar.google.com/scholar_case?case=864795647667642...

This is not legal advice and I am not your lawyer.

Google figured out a magic formula:

1. develop a service (e.g., search)

2. provide the service for free to users

3. sell ads (unobtrusively placed within the service) to marketers

4. profit!

Most companies cannot seem to figure out (or afford to get to) step 3. Google's new services generally muck up the formula. E.g., with Stadia, they've broken the magic formula at both step 2 and step 3:

1. develop Stadia

2. provide Stadia to users for a monthly fee and/or per game fee (not free means no network effect)

3. ????

4. Profit!

> DLSS kinda works because we "know" what is what thing in a photo.

I thought the reason DLSS works is because the same rendering algorithm is used to generate the low resolution image and the high resolution image and the neural network merely learns a filter between the two.

Take a patient with ECoG implant(s), put EEG sensors on the patient, and hit record. You now have the same rendering mechanism (the brain) generating a low resolution signal (EEG) and a high resolution signal (ECoG).

However, back to DLSS, if the low resolution signal is a single pixel, then generating a 4k image from just that single pixel may not be very fruitful.

Still, it would be interesting to see an attempt at using a generative adversarial network (GAN) to generate an ECoG from an EEG. And if it doesn't work, then make a determination of how much more EEG sensitivity is needed before it will work.

Has there been any work to use a neural network to generate/simulate ECoG signal output from an EEG signal input? (My Google-fu only gives definitions and distinctions for ECoG and EEG). Almost sounds similar in concept to deep learning super sampling (DLSS), i.e., taking a low resolution image/signal (EEG) and using a neural network to generate/simulate a high resolution image/signal (ECoG).

Going further, put a small space facing camera on each satellite. The images can be stiched together to improve resolution and maintain continuous observation of space at all angles. Image downloads can be limited to save bandwidth.

My google-fu is failing. I see this and copies of this article, but I want to know more about the I/O available to the AI. Did the AI get full access to the Minecraft API or were they limited audio, video, keyboard, mouse like a human player?

No pics in the SciAm article. The Nature article includes pics:

https://www.nature.com/articles/s41586-019-1687-0

EDIT, NOTE: when clicking from inside the SciAm article, the Nature article was displayed. Clicking on the link from HN shows a paywall.

From the title, I was expecting the neural network to take an input (e.g., speech or a string "5+11+3=") and then control mouse movements to push the keys on a calculator program (e.g., Windows Calculator). I.e., a neural network driving an existing user interface based on commands from a user.

But the article is more about using neural network transformers to build steps of a mathematical proof with each step checked by a symbolic "calculator". I.e., transformers applied to mathematical proofs.

Employees want a say in the governance of the corporation. Most corporations have no mechanism for this leading to impromptu employee protests over the issue of the day.

To have more influence over the governance of a corporation, the employees could push for a bicameral governance structure. E.g., have the board of investors elected directors with the president/CEO running the company, but also create a second board, a board of employee elected directors. The employee elected directors are elected by the employees of the corporation (1 employee = 1 vote). In order for a resolution (e.g., setting c-suite compensation or identifying types of clients the company will not cater to) to pass, both the board of investor elected directors and the board of employee elected directors must pass the resolution.

A benefit to the c-suite of this style may be that the employee elected directors can provide better feedback about the concerns of the workers before the workers start protesting and come up with solutions that are acceptable to both sides.

Air from the atmosphere may be the propellant [1]. If sufficient velocity can be achieved before leaving the atmosphere, then no propellant is needed to be carried up on the way to orbit.

[1] https://www.youtube.com/watch?v=XhUasBcoj-Q

Right, and in response to the parent, a more interesting bet may be whether crewed Dragon flies first or a crewed starship flies first.

Fascinating. Multiple different DNA based instruction set architectures.

https://en.wikipedia.org/wiki/Human_mitochondrial_genetics

Genetic Code Variants

For most organisms the "stop codons" are "UAA", "UAG", and "UGA". In vertebrate mitochondria "AGA" and "AGG" are also stop codons, but not "UGA", which codes for tryptophan instead. "AUA" codes for isoleucine in most organisms but for methionine in vertebrate mitochondrial mRNA.

wikipedia cites to https://www.ncbi.nlm.nih.gov/Taxonomy/Utils/wprintgc.cgi?

Surprised by the lack of an ad supported free version. Seems like applying the lessons they've learned from conquering internet search might be helpful, but apparently not.

Excellent writeup, big thanks.

From the file wrapper (parent post reference [0]), the examiner's main point in the rejection was apparently that the generation of E-fields of 10^18 V/m and B-fields of 10^9 Tesla, as needed for the device, are impossible.

The 2018-10-31 Notice of Allowance states that the application was allowed for the reasons in the 2018-08-21 Appeal brief.

The first part of the appeal brief (in response to the 112a rejection) more or less states that the peer reviewed papers are sufficient to overcome the rejection by proving that electromagnetic flux values of 10^33 Watts/m^2 are possible, which are equivalent to the required E-field and B-field values.

The second part of the appeal brief (in response to the 112b rejection) gives a little more context on the situation with a quote from the AIA 2017-5343 paper (parent post reference [1]):

"It is a well-known facet of quantum field theory that everything can be described in quantum mechanical terms. The complex interactions between a physical system and its surroundings (environment), disrupt the quantum mechanical nature of a system and render it classical under ordinary observation. This process is known as decoherence. However, it is argued that we can retard (delay) decoherence (and possibly even suppress it – namely decouple a physical system from the environment) by accelerated spin and/or accelerated vibration of electrically charged matter under rapid acceleration transients. This may be the very condition to achieve a state of macroscopic quantum coherence, the idea being that we never let the system achieve thermodynamic equilibrium, by constantly delaying the onset of relaxation to equilibrium (hence the production of maximal entropy is delayed). The system may ‘violently’ react by generating ‘anomalous’ emergent phenomena, such as, but not limited to, inertial mass reduction."

The idea of "macroscopic quantum coherence" is fascinating. This could be the next major technological race outside of computing.

[0] https://portal.uspto.gov/pair/PublicPair (search for application number 15/141,270, then click on "Image File Wrapper" tab, then click to select all and download PDF) (includes several other peer reviewed references to support the patent)

[1] https://arc.aiaa.org/doi/pdf/10.2514/6.2017-5343

Sounds like a neural network that converts EEG signals to ECoG signals would be an interesting idea.

If they wanted to monetize it, why not patent the new features, demand royalties from AWS, set royalty rates of $0 for small to mid size businesses, and keep it open source? Or customize the patent so that it's only enforceable against AWS.

I realize "patent" is a dirty word around here, but aside from that, what are the issues with the above strategy and is there truly no possible way around those issues?

Patents weren't the issue. As I recall, Shkreli's company had the exclusive right to market the drug in the US, regardless of patent protection. For example, the FDA provides regulatory exclusivity that prevents generics from entering the market even after the patent expires. [1]

Patents were a collateral issue. On that collateral issue, one of the issues with patents is that drug companies have a tendency to stop producing drugs towards the end of life of a patent to force patients onto a newly patented drug. This practice could be stopped by allowing generics to produce, market, and sell when the patent holder will not.

[1] https://www.commonwealthfund.org/publications/journal-articl...

When Shkreli got vilified, he missed out on one of the biggest opportunities to push for meaningful change. Media and Congress were "not kind" to his price increases. Well, who passed the laws that allowed for those price increases and stifled competition? At the time, I was hoping Shkreli would double down, go all in, and blame Congress for passing the laws that were allowing for such price increases and preventing competition (which big pharma had lobbied for and gotten passed earlier). Demand that if Congress was serious about getting him to reduce prices, all they had to do was repeal those laws. I.e., pass the buck back and say "well if you don't like this behavior, then why did you pass these laws?" and "oh, by the way, how many campaign contributions and PAC donations did you get for passing these laws?" He could then pivot to say "hey, I'm not raising prices to make money [at least not anymore], I'm raising prices to increase awareness and bring accountability to Congress and hold their feet to the fire."

Emotional Intelligence 2.0 by Travis Bradberry. Two people had to recommend it to me independently before I finally read it. Reading and deeply contemplating this book dramatically increased my ability to understand and predict emotionally driven behaviors. Drama in movies and in life started to make sense.

> Market share tax just results in companies redefining their industry.

Do you have a specific example?

But how to "Fragment the market legally"? Breakups lead to more monopolies [1]. Could the tax code be manipulated to achieve the desired effect? A market share tax could tax revenue of large corporations (revenue >= $1B [2]) on products and services that are over X% market share at Y%. With X=40% and Y=80%, then:

$4.2B revenue with 40% market share has $0 market share tax and keeps $4.2B; $5B revenue with 50% market share has $800M market share tax and keeps $4.2B; $7B revenue with 80% market share has $2.8B market share tax and keeps $4.2B;

This would reduce the incentive to grow market share above the threshold. A 40% threshold allows for two companies to max out market share and a 30% threshold allows for three companies to max out market share.

[1] https://en.wikipedia.org/wiki/Breakup_of_the_Bell_System [2] https://www.gartner.com/it-glossary/smbs-small-and-midsize-b...