psandersen
If someone believes consciousness arises from a non-physical spirit through an unknown mechanism; then it seems difficult to argue that this can only happen in the specific arrangements of biological cells that we've seen on earth and not in other ways.
E.g. Donald Hoffman's view around consciousness being fundamental certainly wouldn't preclude a machine operating as a portal* in the same way humans do.
*I'm only familiar with his ideas and the concept of 'portals' from his recent appearance on the Lex Fridman and TOE podcasts.
There are many steps to move a democracy to a totalitarian state, and the time to hit the breaks is at earliest signs of fascist behavior.
Abstractly, the people who excused and weakened institutions may have done more to create an eventual totalitarian state than the asshole who moved the final 1% and seized power.
I'd primarily like to be able to turn on/off a few specific switches, set a few specific scenes and ideally play a few playlists (squeezebox integrated with home assistant). Would creating scripts in HA and then calling them from Hi be the way to go for customising without building custom components?
If this works well I'll be keen to add a few cheap pis plus ps3 eye cameras around the house.
Also if you wouldn't mind elaborating on surprising ways reaction diffusion equations can be used to solve problems I'd love to read it :)
IMHO the core problem is the recommendation engines are optimizing the wrong metrics and this leads to destabilizing feedback loops.
The problem is likely AI complete, but it should be possible to tune recommendations so they don't create self strengthening conspiracy theories. I suspect a reinforcement learning environment that mimics users interacting with a recommender system would be the best way of tuning the larger system behavior so the longer term properties aren't catastrophic for society.
OTOH if the platform was setup so "good faith" arguments were somehow incentivized, there might not be any need to optimize for any other metric than engagement.
IMHO the best approach if you suspect you may have ADHD is to do a lot of googling for adult ADHD, there are good videos on youtube, discussions on reddit and various forums. If you still think this might apply to you, then search for ADHD test.
If you have ADHD there is a rich set of literature on managing it, supplements to investigate (lots of discussions in the nootropic community) and the meds can be a life changing tool when used responsibly.
I found it compelling and switched, feeling healthier and lost around 20kg in 6 months without thinking about calories.
That would indeed be an optimal compressor and start to enable things like Solomonof induction and AIXI.
Seems the challenge is to get feedback to guide the search for the right Turing machine. With infinite compute can just run an ever expanding set of Turing machines until you get one that generates the input, but without some breakthrough that would enable something like stochastic gradient descent to efficiently work on non-differentiable programs I don't see how it can be made practical.
Perhaps there is some sweet spot thats more general than current compressors that looks at a restrict class of Turing machines that could be feasible.
Exploring a pocket to pdf to kindle pipeline is on my todolist using either https://www.printfriendly.com or https://www.api2pdf.com
If anyone has done something similar would love to some feedback.
Maybe this https://www.coursera.org/learn/automated-reasoning-sat would be a good start?
Ideally I would like to be able to take predicted outputs from ML plus business requirements to solve allocation problems; and more pie in the sky I would like to know enough to get inspiration for how to link a neural network with a sat solver, such as perhaps replacing or augmenting beam search in sequence decoding tasks.
I meant that the splitting point in a node in the decision trees that make up a random forest is itself a random variable that follows a distribution. Because its a smooth function (i.e. probability of splitting is 50% at the point and rises/falls smoothly) it should in principle be differentiable* so that the whole model can be trained by SGD and/or fit into an end to end learning pipeline with convolutional layers etc.
*Where I'm hazy, is how can a smooth probability function be differentiable when sampling. I'm brainstorming in the open here, will do some reading on stochastic neural networks.
I've been thinking about extensions to decision tree models that could get the benefits of NNs and it seems like there are a few ideas floating around.
For example; Probabilistic Random Forests have some really interesting properties for noisy datasets, e.g. "The PRF accuracy decreased by less then 5% for a dataset with as many as 45% misclassified objects, compared to a clean dataset." - https://arxiv.org/abs/1811.05994
PRF's might be a natural fit for RL, especially methods using monte carlo tree search.
Speculating here as I'm not adequately familiar with stochastic calculus, but intuitively it seems like probabilistic decision trees could be made differentiable since the hard decision threshold in a tree could be a turned continuous (i.e. every split is logistic regression), which might enable some really interesting applications. I personally dream of being able to cleanly integrate decision trees and tools from NNs in something like pyro for a fully Bayesian model.
I have a respectable (2x8 core xeon, 164gb ram) workstation running linux and I have many tabs/windows open with various references spread across virtual desktops optimised for tasks (work project 1, 2, education, personal project,... etc).
I've experimented with various extensions that pause tabs and currently use tab wrangler to auto-close tabs but it still gets slower to respond with ~50 tabs open across all windows. Sometimes I check the firefox & system task managers to see some basic reference tab (i.e. blog post) taking most of the CPU...
It would be great to be able to aggressively throttle inactive tabs or even pause them completely so only ram is used.
Some options around this use case would be much appreciated. Responsiveness under (almost) all conditions is much more important than raw benchmark throughput. Near instant tab resuming would be a huge help towards this goal as well.
Its hard for us with a very mixed tech stack to realise something like this. I've started thinking about using VNC to share Autodesk Sketchbook from my Surface Pro to do presentations and splicing it together with my webcam using OBS for meetings.
If anyone has done something similar I'd love to hear it, main workstation is linux.
Is there any chance we could get it to make a rest API call on certain actions? I'd like to change my lights in home assistant as an added visual reinforcement.
Causal forests are an example that extends random forests, but I imagine a lot of the value in current pipelines would be to use causality as regulariser. This could be a parameter that controls the weight of established causal links, or it could be as a scaffold; e.g. a first 'causal pass' is used to establish constraints (monotonicity, conditional variable selection, reject changes that result in predictions inconsistent with the initial causal model when there is a strong causal model etc).
RL is likely more promising. If agents could be made to search for causality in an environment these relationships could be made much harder to unlearn which would then enable more efficient exploration & incremental learning. Framed this way causality guides attention, limits the search space and locks in learning.
I've got some quarantine reading/experiments to try! :)
That said, I can really relate to the challenge of keeping personal and professional separate when working from home... I keep thinking of setting up a separate room or taking a laptop outside but it is painful giving up comfortable the multi-monitor setup on my main workstation.
To add additional detail, I'm currently using OpenMTCPRouter to bond 4G and DSL but it suffers when one connection starts experiencing packet loss.
There are plenty of guides out there that use shadowsocks which works well for web-browsing, but to be truly transparent need to just set a gateway to route via the tunnel. A docker image or a vm image that makes needs minimal config on both ends would be ideal. Has anyone done something similar?
It would be pretty unethical to prevent people from getting the best available medical treatment just so there are more children, on the other hand giving people a choice would have many social implications.
If the supplier is only passing on their own increased costs then that's at least fair in principle. Similarly, marginal increases in price for marginal increases in use also seems fair when there are demand surges.
This illustrates the danger of ranking everything relatively... if enough people watch infowars like content then every university will be marked as radical left wing etc. There needs to be some way of anchoring the spectrums in actual facts or we're firmly in the 'post truth' era.
Features likely include more than just the single users history, so they need to be updated often enough for the model to do fast predictions. E.g. you want your model to capture if many people are booking from the same area at once because there were results from a sports game etc, but you dont want to run an expensive query for every user of the page.
Definitely not the first thing to worry about in a startup, but better performance at Booking.com's scale is serious $$$.