If anyone builds a version of this that accepts crowd sourced phone images to increase the accuracy with photogrammetry (before I get around to it) I will give you shademaps.com.
> Similar[ly] the prevalence of other anatomical features such as spina bifida occulta and fabella has increased over the last 2–3 centuries. Evidence indicates that changes in the natural selection pressures acting on these specific anatomical features could have caused microevolutionary processes, leading to the observed increases in prevalence rates.
They seem to be suggesting that this is not necessarily an advantage. Modern medicine makes pregnancies with these genetic mutations more viable, so the dominance of the gene expression becomes a more significant factor than fitness.
I have been working on a portable camera array to put real video on these displays. The video demos are either projected depth maps or very expensive, stationary, indoor camera rigs. Despite trying to stay away from AI generation, making the array sparse, then synthesizing missing angles and dropped frames actually solved a lot of problems better than more hardware.
There is a much older Taco Bell grift: the charity coin drop game. Drop a coin onto the bottom step of an acrylic spiral staircase in a little water tank and you get a free menu item based on the denomination of the coin. It was designed so that if you landed on the last 2-3 steps you had a chance, but any higher and the centrifugal force of walking it down the steps would send the coin off the edge. It was very random with poor odds.
But eventually employees got sick of changing nasty change water and left it empty. The physics were a little different, but it seemed to be sufficiently difficult. Rather than wafting around, the further the coin dropped the more likely it was to bounce off a step. The much more deterministic coin movement enabled a new strategy: walk the coin backwards off of the first step and catch the coin directly on the final step. After ~$1 of practice, I had years of 5¢ tacos.
You can get a tutor to check your homework and highlight your errors. Sometimes they just tell you answers in the process of teaching. AI can be a tutor if that is what you ask for.
Homework is for using every tool at your disposal to learn. Tests are for proving you actually learned. Grading homework for credit and not giving feedback until after the test is a pattern I have seen too many times in higher ed, and it is a sorry excuse for education.
It looks like it could be printed flat and they are forcing everything in one print. That also increases the chance of failure and maximizes the resulting waste. I have wasted enough material with long multipart prints that I limit myself to ~8h if I can.
It would be nice if articles would cite a source other than twitter commentary. The official verdict makes it clear this was about operating a laundering service not publishing a laundering tool.
> The relayer acts as an intermediary and arranges the full withdrawal, including the payment of the fee. … the transaction track of the cryptocurrency in question stops at Tornado Cash. …
> When these acts are carried out in respect of crime-derived Ether, it is in fact Tornado Cash that carries out the concealing or concealing money laundering act. Therefore, in the opinion of the court, Tornado Cash cannot be seen merely as a tool for the user.
I don’t understand the use case. You go through the trouble of generating checksums when copying videos, but don’t want to regenerate the checksums when modifying the metadata? If you are this concerned about data corruption why not check the metadata also?
There is good open source cutter plotting software available, but Cricut sues everyone who bypasses their app. I suspect their hardware is being subsidized by these asinine paywalls to monopolize the consumer cutter market.
I find that ingesting and chunking PDF textbooks automatically creates more of a fuzzy keyword index than a high level conceptual knowledge base. Manually curating the text into chunks and annotating high level context is an improvement, but it seems like chunks should be stored as a dependency tree so that, regardless of delineation, on retrieval the full context is recovered.
The account management / status tools were slow and flaky on the best days. I wouldn’t rule out a little extra traffic knocking them out. Correlation is not necessarily causation.
I ran into a GitHub issue template a long time ago that had a required section for alternate solutions. It’s amazing how often you find a better solution when you are forced to enumerate viable solutions and compare them.
My heuristic is how much noise is in the closest vectors. Even if the top k matches seem good, if the following noise has practically identical distance scores, it is going to fail a lot in practice. Ideally you could calculate some constant threshold so that everything closer is relevant and everything further is irrelevant.
It is two different language models. The embedding model tries to capture too many irrelevant aspects of the prompt that ends up putting it close to seemingly random documents. Inverting the question into the LLM’s blind guess and distilling it down to keywords causes the embedding to be very sparse and specific. A popular strategy has been to invert the documents into questions during initial embedding, but I think that is a performance hack that still suffers from sentence prompts being bad vector indexes.
Instead of embedding the user prompt, I let the LLM invert it into keywords and search the embedding of that. It very much does feel like a magic bullet.
My LG OLED had intermittent contrast issues that persisted after toggling every permutation of picture bell and whistle. After finally stumbling across the correct search term, it turned out the offending feature was hidden in a maintenance menu that requires a button that only exists on service remotes… The same issue was fixed in an update for newer models. Tons of forum posts on the subject resulted in RMAs and refunds. How did we get like this?
Instagram is the same. They steal video from real products, funnel purchases through a random e-commerce page, and disappear by the time you need to request a charge back so that you can’t find evidence of the original offer.
The preview image shows intellisense covering up 80% of the code with irrelevant suggestions and a compiler error verbosely complaining about code in progress. That much distraction should have a very small visual indicator at the point of focus, just enough to hint the action required to expand if it is actually relevant.
The assertion that DHS was funding the program (or using it as a tool) is an unsubstantiated leap. Stanford was gathering and analyzing intelligence with national security implications. There is an obvious public service in sharing this information with DHS.
It is misrepresenting the facts to the extent of being misinformation, which is quite ironic. It builds a grand narrative of conspiracy, but when the substantiating facts are finally sprinkled in they don’t quite match.
> we just set up an election integrity partnership at the request of DHS/CISA
> Brookie’s acknowledgment contradicts the claim made by the EIP that the idea for the project “came from four students
The partnership was not the program. This glaring misrepresentation is what the entire piece is built on.
Blu-ray is not raw, it is just HEVC with a higher bitrate. Most content encoded at a high profile level is not even going to hit the bitrate limit of a good streaming service + connection. Other than pausing during an action scene to take a still, I can’t perceive the difference.
I would imagine the reliability would be higher. A resistive heating element could degrade and eventually develop a fault, but an inductive circuit decouples the electrical circuit from the heating element. A fault or an impurity in the heating material wouldn’t prevent the inductive field from exciting the surrounding conductor particles.
The Dolby Atmos/Vision standards also provide HDR, dynamic brightness, and surround sound. People spend thousands of dollars on home theater setups to meet those standards. Max has almost completely stopped producing cinematic content, so I can’t imagine the upsell making much sense.