204 karma · joined November 28, 2016
Rigging also looks like it could have a decent AI/DNN solution: https://arxiv.org/pdf/2005.00559.pdf
The article below estimates the Shahed-136 at $20,000. The Mohajer-6 probably isn't that much more.
https://www.timesofisrael.com/the-iranian-made-killer-drones...
But that's the economic-warfare genius of cheap drones: it [might] cost as much $500,000 to shoot down a $20,000 drone.
https://www.businessinsider.com/suicide-drones-much-cheaper-...
I don't need to agree with every expert/leader, I just want to know what they think.
The robots do not seem specifically involved with removing spoil or even carving it into smaller pieces (at least pieces smaller than the tunnel diameter).
Does this mean building tunnel shells is the hardest part of tunnel building, not breaking up and removing spoil?
Or does this approach somehow make it easier to fragment and remove the spoil since it's now in tunnel-sized chunks?
(Edit: Yes, stabilising the tunnel is the hard part, see excellent and thorough answer by rmccue)
> While ChatGPT is able to remember what the user has said earlier in the conversation, there is a limit to how much information it can retain. The model is able to reference up to approximately 3000 words (or 4000 tokens) from the current conversation - any information beyond that is not stored.
This implies ChatGPT has a 4000 token maximum prompt and prior prompts in a given web session are inserted into the current prompt, most recent to oldest (probably with some sort of time context like "previously, user asked:"), up to 4000 tokens.
In the essay, identity is associated with a particular religious/political/social perspective on the human condition, not the human condition itself. Keeping identity small is basically converging on "being human"-- the choices everyone must make as you note -- without the false certainty that comes with loyalty to a religious/political/social identity.
https://www.flir.com/news-center/camera-cores--components/fl...
"ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this issue is challenging, as: (1) during RL training, there’s currently no source of truth; (2) training the model to be more cautious causes it to decline questions that it can answer correctly; and (3) supervised training misleads the model because the ideal answer depends on what the model knows, rather than what the human demonstrator knows."
https://transcripts.cnn.com/show/fzgps/date/2022-07-31/segme...
LIU: Nobody can control TSMC by force. If you take a military force or invasion you will render TSMC's factory non-operable because this is such a sophisticated manufacturing facility. It depends on the real time [correlation] with the outside world, with Europe, with Japan, with the U.S., from materials to chemicals to spare parts to engineering software diagnosis. And it's everybody's effort to maybe this factory operable. So if you take it over by force you can no longer make it operable.
"This study had several limitations. First, our sampling strategy selected for members of Santa Clara County with access to Facebook and a car to attend drive-through testing sites. This resulted in an overrepresentation of white women between the ages of 19 and 64, and an under-representation of Hispanic and Asian populations, relative to our community. Those imbalances were partly addressed by weighting our sample population by zip code, race, and sex to match the county. We did not account for age imbalance in our sample, and could not ascertain representativeness of SARS-CoV-2 antibodies in homeless populations. Other biases, such as bias favoring individuals in good health capable of attending our testing sites, or bias favoring those with prior COVID-like illnesses seeking antibody confirmation are also possible. The overall effect of such biases is hard to ascertain."
"We used Facebook to quickly reach a large number of county residents and because it allows for granular targeting by zip code and sociodemographic characteristics. We used a combination of two targeting strategies: ads aimed at a representative population of the county by zip code, and specially targeted ads to balance our sample for under-represented zip codes. In addition, we capped registrations from overrepresented areas."
Update: Okay, "Other biases, such as bias favoring individuals in good health capable of attending our testing sites, or bias favoring those with prior COVID-like illnesses seeking antibody confirmation are also possible."
https://www.journalofinfection.com/article/S0163-4453(20)301...
Further, the internal monologue is just one representation in mind, it does not replace the full (non-verbalized) thought. Both are present and complement each other.
Facial recognition in the brain is just specialized object recognition, therefore any research on powerful general-purpose object recognition via deep-neural-networks is improving facial recognition technology whether it means to or not. Only a law banning all artificial neural network research seems to have a chance at stopping the technical reality here.
Edit: Okay, remote driver access might solve this.
> Of all the anti-aging and longevity-promoting interventions tested so far, calorie restriction (CR) seems to be the most conserved among species and robust. However, humans struggle to constantly achieve a certain level of CR. Here, we show in a clinical trial that a related intervention, alternate day fasting (ADF), also leads to striking reduction in overall calorie intake over the course of the study but is more easily tolerated than continuous CR and provokes similar beneficial changes on the cardiovascular system and on body composition while being safe for a period of >6 months.
“The technology has been developed through a combination of supervised and unsupervised learning techniques. Unsupervised learning was used for teaching the AI engine the core legalese language. Thereafter, supervised learning, using deep learning multi-layer LSTM and convolution technology, was used to train the system for the fine-tuned issue-spotting. Supervision was performed based on human-annotated documents, using legal experts. A unique augmentation algorithm was applied to boost learning from these examples. The overall result is the most advanced technology for the automatic analysis of legal documents. The p-value for the statement that accuracy of AI is above that of these lawyers is 0.0068 (using MannWhitney’s U test).
https://images.law.com/contrib/content/uploads/documents/397...
So, clearly most Americans don't stay behind when their town has no future. The article should add the qualifier "Some" Americans to avoid a misleading impression.
If evolution can be taken as a successful search through the space of what is possible, language and reason are on the path from curve fitting. But calling this "easy" is far from my point. If evolution is any sort of guide, deep neural network architectures are where the real work is needed before language understanding is possible (contra Pearl, Marcus).
Homo neanderthalesis or thereabouts to modern day: language, reason, cause and effect.
> Today’s dominant approach to A.I. has not worked out.
But "curve fitting" biological neural networks needed 2500 times more architectural exploration and enhancement to get to rudimentary language capacity than rudimentary language architectures needed to achieve all of human knowledge. That suggests we're doing the right thing in growing and enhancing deep neural network architectures now. Evolution suggests the next step to language and reason on the right architecture is comparatively easy.
So is the OP:
"We’re releasing an analysis showing that since 2012, the amount of compute [amount of calculation, amount of computation] used in the largest AI training runs has been increasing exponentially with a 3.5 month-doubling time "
Without loss of meaning, title could be AI and Calculation, or AI and Computation