Although the article focused primarily on AlphaFold, many other ML approaches are making impactful contributions in the general scientific field. One example is the diffusion model and its use of stochastic differential equations (SDEs).
I personally dislike these kinds of comments. The idea that androgens aid in recovery may be known in theory, but this study provides empirical evidence to support that claim. So yes, there is scientific value in it.
> Anabolic steroids may aid in the healing of muscle contusion injury to speed the recovery of force-generating capacity. Although anabolic steroids are considered renegade drugs, they may have an ethical clinical application to aid healing in severe muscle contusion injury, and their use in the treatment of muscle injuries warrants further research.
Perhaps. But I don't feel they're saying "in theory". They're saying we've already seen that XX and XY recover differently, and this is simply another example of something we already know.
Sure a link or two might have helped. But there is value in pointing out "no new news here" if that provides context.
economists proudly ignore real world data. They care about their models, even when they are obviously wrong.
Milton Friedman said "Truly important and significant hypotheses will be found to have 'assumptions' that are wildly inaccurate descriptive representations of reality, and, in general, the more significant the theory, the more unrealistic the assumptions."
Distribution shift in the real world data will always be inherent to any data driven methods. Unless there are major advances in continual learning for DL models, they will always struggle with distribution shift degradation.
Similarly, humans are also prone to the distribution shift unless we get updated information on a specific topic. The key differences are that we are great at continual learning and we are much better at learning abstraction
A 17 million parameter model (~Resnet50) takes more than 50s proof time. Is this on top of the inference time?
I can see some niche applications for this system, but I am very skeptical it's ability to handle larger models (100M+) and the ability to and it's scalability when there are increased demand.
It's not impossible to see that LLMs can make finite step "reasoning" between input and output, as each block of the transformer can model probabilistic causality. Transformers are sometimes considered a fully connected graph neural network, which can be used for modeling causal graphs. One additional supporting evidence on the finite step reasoning hypothesis is that "train of thought" improves the performance of LLMs[1], meaning that letting the model explain itself explicitly reduces the amount of implicit "reasoning" steps that need to happen within the model.
Additionally, while we don't know how us humans functionally reason, it's believed that the predictive nature of our brain is central to our reasoning abilities. Maybe in some way, the autoregressive nature of LLMs is similar to our predictive brain.
For a collection of emergent properties in LLMs, I recommend this paper [2]
Apple Watch has so many health features, but the main hinderance for me is its relatively short battery life - I often forget to put it back on after charging it.
Has anyone tried having two apple watches so one can be worn while the other charges?
[edit] Charging it at night is kind of missing the point, as I would like to track both my daily activities and my sleep
> Has anyone tried having two apple watches so one can be worn while the other charges?
It's terrible that it's come to this IMO.
The Apple watch has some great features and can't directly compare, but I can't help compare it to Pebble watches almost 10 years ago - weeks worth of battery life. Even when it "died" it would work for days longer as a watch.
I know the Apple watch has different features, constant heart rate and accelerometer monitoring, etc - but how are we supposed to use it for sleep monitoring (my usecase) if we need to charge it at night anyway? I find it's really, really good at it and I want to use it so badly.
Surely this is a stupidly huge flaw if the battery life is so low that people are considering a second watch just for monitoring the second half of their day (sleeping)?
(I still own an apple watch, and I know there's still a large community supporting them but I wish they were still manufactured)
This may be true for older Apple Watches, but it's not true for the latest Apple Watch. I get 3 days of battery life on regular usage, and a full charge only takes about an hour. Throwing it on the charger while I take a shower or wash dishes is enough to keep it topped off.
Also Apple introduced a "Low Power Mode" on WatchOS 9 that allows you to potentially get 60 hrs on the Apple watch ultra. This link discussed what is disabled in this mode. Seems like it would work for your use case or at the very least you run low power mode during the day and turn it off at night for better sleep monitoring: https://www.macrumors.com/how-to/enable-low-power-mode-apple...
I have a series 5 and it easily last 24 hours, with the screen on the entire time. It charges pretty fast. Haven't measured it but I never have to think about battery life.
Even older ones are decent. I have a 4 and it's probably got about 2 days of battery life. I put it on the charger for an hour every morning as part of my routine.
The first thing I did was turn-off the always on display, which save a lot of energy.
I also turned off the background processing of most apps, which saves battery. Same thing with location if the app doesn’t require it to function properly.
My friends with the newer watches/chargers say that they throw it on the charger when they get up to hit the shower in the morning and its mostly charged by the time they're ready to head out to work.
I've done that have two apple watches I used to have a series 7 and a se charge. But I've gotten an Ultra and I only have to charge it for a couple minutes to get a whole night worth of tracking.
I usually charge my watch while sleeping and grab it in the morning along with my phone, but I'd assume a second watch would work? It's an expensive route with the watch, but it's a solution in many other spaces. My wireless headset, for example, has a replaceable, rechargeable battery.
For best results, you'd want to use a single charger, and to always put the alternate on immediately.
That said, the watch can notify when charged, which might be a better first step. The settings for that live in the sleep settings, of all places, but might be worth a shot?
I just got the Apple Watch 8, it's my first Apple Watch. The battery life is excellent. Lasts for 3 days from a full charge. Throwing it on the charger for a few minutes every day while I wash dishes is enough to keep it charged indefinitely.
VPN might have a part in this too. I remember when I was in China, I was connected to ExpressVPN 24/7 (whatsapp, gmail etc), and the battery drain was visibly higher than when I had the VPN off.
While this may be amongst the long term goals, I don't think it's useful for military applications yet until laser com is fully online. Currently there is no sat to sat com, so there's is always ground station within sight of a sat. Unless SpaceX places a ground station near a conflict zone, military can't utilize it much yet.
I believe this applies to all currencies - they don't have intrinsic values. The values are derived from the market. Be it USD, BTC, EUR, or the Zimbabwean dollar. However, the purpose of currency exists as a vehicle for wealth transfer.
IMO the biggest problem with most cryptocurrencies is their deflationary nature. Limited supply means they behave more like assets than currency. As currency nobody would want to spend it as they gain value over time. As an asset it lacks intrinsic value, and therefore overall market cap will tank slowly.
> As currency nobody would want to spend it as they gain value over time
The problem with this argument is that you have to spend money. You have to spend it on food, lodging, entertainment, and so forth. Even if a currency is deflationary, people still have needs.
Yes, but one would prefer spending money using an inflationary currency since holding on to it devalues over time. As long as inflationary currencies exist, people will prefer to spending those first before deflationary currencies.
Let's say you store all your wealth in a deflationary currency, for the obvious reason that it gains value over time. Then converting it to an inflationary currency would simply add overhead. So you would spend the deflationary currency.
Your opening hypothetical leaves the impression that the words which follow it were either an alternate rehearsal for a more blunt observation: "if all your wealth is deflationary, to spend is to lose deflationary currency" or an earnest counterargument-by-analogy which contradicts something very near the bedrock of my own common sense about money; why would anyone with savings in a deflationary currency and the opportunity to spend inflationary currency decide to have & hold that which would decrease in value in favor of (equivalently denominated) spending which won't?
I suggest you look into modern economic theories and why inflation is crucial to economic development. There are two problematic assumptions in your statement:
1. Nobody would store all their wealth in a deflationary currency - other forms of assets, stocks etc will still exist. They will need a high liquidity vehicle to trade a.k.a inflationary currency.
2. Deflationary currency doesn't necessarily mean it will gain value overtime. Unless it contains growing intrinsic value, the overall market cap will drop/plateau.
Think of inflation vs deflation as lubricant vs sandpaper, one encourages economic activities while the other discourages.
1. bitcoin is high liquidity (even higher liquidity than currencies in that you can trivially transfer $1B), unlike all of the other assets like stocks and houses....
2. Exactly which is why people would be fine spending a "deflationary" currency
I believe this is completely wrong, the values of many important currencies are not at all derived from some vague market value, but are very actively centrally fought for, and controlled by the respective governments, and their respective economic and military might, and reserves - no equivalent in the typical cryptocurrency world. The decision to devalue the currency to save the economy, the decision to change course and peg a currency to something else entirely, the centralized fight to contain hyperinflation with all resources of a government, these have nothing to do with "deriving values from the market".
In contrast, there is no central government with military and reserves that can decide to manufacture an extra $3 trillion same price BTC on a whim to temporarily address a virus, or to devalue BTC to save the local economy, most crypto "currencies" just don't seem to work that way.
You didn't quite get the point. As individual currencies governments exert a lot of power over their relative value to other currencies. As a whole, worlds fiat currencies hold no intrinsic value. No government can dictate that a kilogram of gold to only cost $1.
Microsoft has an initiative called AI4Sciencie (https://www.microsoft.com/en-us/research/lab/microsoft-resea...) which published a fair amount of SDE/diffusion-based method to solve scientific problems