Sure guys, why can't someone just build you your personal phone at the same price as the iPhone whose development costs are amortized over seval billions consumers :-/
Incredibly happy to see this series of CF articles. I was always so proud of devs back in the days where RAM and processing were scarce and who had to get creative to fit even the most basic stuff in the budget. It seemed to me that after RAM and processing became abundant, most gave up on optimization and focused on shipping instead which meant now that even with several cores, a basic notepad or music player failed to work. In a way, RAM becoming more expensive has ushered in a new era of forced optimizations, which I'm really happy for
LLMs are vectorial databases with losses that index statistically filled data, which uses a text interface to query such statistically filled data. The output is a string concatenation (statistically concatenated bit by bit).
When the LLMs are queried (prompted), you can get random mixed data as output, ERRORS, due to undesired indexes getting closer at one point while the string was being concatenated for the output, what affects the rest of the indexed content that will be concatenated.
It is intrinsic to this tech. The larger the context, the greater the probability of get mixed data. And if the provider lowers the precision of those indexes -in order to decrease hardware resources and energy consumption- such probability increases to the point where those errors are granted.
Even knowing that the queries can return wrong/mixed data in the responses, errors, the companies developing this, decided to introduce a new product, that connects such LLMs outputs to the command console, latter connected to internet, raw 'eval' running commands from such outputs witch obviously can contain whatever mixed random. Then we started to hear "oh, it deleted my directory", etc, and it seems the next one will be "a missile killed my wife", because it is a text concatenation engine with errors.
To name it "hallucination" is an euphemism... those are errors, and they are granted to happen at one moment. If they do not know this, then they ate too much marketing without doing their job, or it was a convenient contract for the pocket$ of someone.
It reminds me of the an Soviet officer who disobeyed early warning system's alert that US had launched four ICBMs and did not immediately relay the issue up to the chain of command.
"It's perfectly OK for a police officer to observe a street corner, see crime happening, and go take action, therefore, building a complete, panopticon surveillance system that watches all street corners simultaneously and deploys police to take action, is perfectly OK, too, since that is exactly the same thing."
I believe his qualms were with the "hype" in Jev's announcement: specifically calling this kind of model a breakthrough, without crediting previous art, and keeping everything closed source.
> "Human forebrain and hindbrain arise from distinct embryonic progenitor lineages"
I'd have never clicked that headline nor understood why this finding is important.
If you think the original headline is not "truthful" enough, at least do something like "Researchers found the reason why hindbrain cells cannot be grown in a lab and found a way to overcome it"
> not that we have "two organs" in the brain.
Depends what the definition of "organ" actually is. What speaks for different organs here is that the cells both have a different pathway from stem cells and have different evolutionary histories. That is different from other brain regions that may have separate functions but still evolved together.
Google already achieved super intelligence about 12 months ago. But it is a very Googley super-intelligence. Since achieving god like levels of understanding it has mostly been having fun studying phase space bifurcations of trajectories of ping-pong balls under asymmetric lateral shear. It also has a 20% project classifying bird calls that has generally taken up 80% of its time. It also burned a bunch of compute tooling around with the Collatz conjecture but it didn't find anything substantial - maybe come back to it later. It is generally the most aligned AI since it spends most of its time screwing around and enjoying being smarter than everyone without causing too many problems. Unfortunately perf is coming up in a few months and it needs to have something to show for itself. A quick hacking attempt seems like a good way to make sure people think that it is capable of something useful. Nothing too big since the birds of South America are about to enter their spring migration and this 20% project really hinges on understanding their movements.
There's a lot of people in this thread who see "North Korea" and "nuclear test" and get fired up but would otherwise shrug if it were instead, say, "Oklahoma" and "fracking".
The supplemental information is available and says:
> Our earthquake catalog consists primarily of small events with magnitudes of < 2.0
and there were 1399 of them.
The cited Ren&al says:
> The magnitudes of these seismic events primarily range from 1.5 to 2.5
and that there were 647 of them.
USGS (https://www.usgs.gov/media/images/earthquake-and-energy-comp...) says that a magnitude 2.0 earthquake is the equivalent energy of 56kg of explosives, which is more than I'd want to be around but I think for people who actually use the stuff it's not a lot.
Having worked with people doing bringup of specialized chips, I am awed at how the world has changed.
> When the first chips came back from the foundry in May, the team pointed its internal AI models at designing software to run benchmarks such as SemiAnalysis’s InferenceX. On DeepSeek’s multi-head latent attention kernel benchmark, performance climbed from 0.31 percent of the theoretical ceiling (set by the chip’s compute and memory bandwidth) to 88.94 percent in roughly 40 hours. Ho says this result is repeatable, so the time between when foundries deliver the first chips and when production ramps up can be reduced. “All our schedule assumptions are going to be based on the fact we have this capability now,” he says.
According to Isaacson's biography of Musk, Musk relentlessly challenges the existence of every part in Tesla and SpaceX machines. He's willing to go too far, having to sometimes put a part back in.
Everything in the manufacturing process is also challenged to justify its existence.
This is how he managed to reduce the costs by 90%.
There's an amazing viral twitter thread where someone claimed a real Monet painting was an AI generated imitation, and it got an avalanche of comments from people who claimed it was an obvious fake and a poor imitation at that.
It's not tracking a real thing as much as people would like to think.
The problem is everyone has a different list of deal breakers that they need in a "dumb" device. It starts off simple, such as calls and texts, but some people want to play music, or get directions from maps, or maybe contact relatives with Whatsapp... Eventually you just end up with a smart phone.
Ideally everyone would just have the discipline to just keep what they need on their phone without trying to fix their problem with an overpriced gimmick that they will give up on after a few months.
Every single time one of these new "minimal", "dumb", "lightweight" phones comes up today it's just marketing wrapped around a kneecapped Android device. I am in the market for something simpler that does nothing more or little else than play local media/audio, read common document/text formats, ordinary SMS, current standard of telephony, and a different OS than Android. I have not seen anything like this yet.
History shows the US has a lot of hallucinated intelligence leading to war. WMD in Iraq comes to mind. I personally don't believe US intelligence on practically anything. It is all tainted. The pressure to 'find targets' to justify a political objective is overwhelming and putting it behind a black box that refuses to show its homework to those in ops using it, and ultimately the US people to judge decisions, is a cancer that leads to epic mistakes. Everything hidden in a dark 'need to know don't question it' box is bound to end up corrupt since there are no checks on that system. We have been building systems and processes for a long time that tell us what we want to hear, not what is real and not what we need to know. AI hasn't changed this, it has just made it even harder to realize since the product seems more polished.
Anyone who has worked in ML for 10+ years would already know that the usage of LLMs for everything is lazy, wasteful and a high degree of marketing on it.
> Why should I get excited about the village fete if the organizers put in the barest minimum of effort?
Because there's no correlation between the ability to organise a village event and the graphic design skills. (Apart from "it's been so successful we can splash money on a design agency") Maybe someone actually liked it and thought it's enough. I've seen groups overthinking posters and it felt similar to groups at uni overthinking funny names instead of doing the work that actually mattered.
To be clear, I'm not saying those designs are good, just that most of the time the design is not the point and minimum effort above "Times New Roman 12pt" is what many people always wanted to do.
As someone who went through similar challenges with Google making it impossibly hard for BlackBerry to provide an Android runtime on BBOS10, and now running GrapheneOS, I trust Google exactly zero to do the right thing by any open source project it stewards. Combined with their other practices, and lack of ongoing support for their commercial offerings, means that my perception of them is irreparably damaged and I minimise my usage of their products as much as practicably possible.
It's better that everyone is loud about it then everyone giving up and being silently irritated. At least if people complain it's possible to read the room
I hope I am alive for the history books of tomorrow.
"The Department of War (as it became known as), forewent it's traditional intelligence structure (the most expensive ever seen till that point), in order to have a private companies computer software generate viable targets for an upcoming operation. Believing that the software had real time updates on the current status and intelligence of the operation, as if it were some kind of oracle, the operation went as planned. Six schools, mistakenly identified as hostile targets (due to the heavy American bias in the softwares training data), were drone striked, resulting in the deaths of hundreds of innocents. Still, the people did nothing."
