mustang Posted November 24, 2023 Author Posted November 24, 2023 (edited) Utefterisi negde, please Ne znamo samo na koje gluposti misliš? Tesla i spacex su prilicno profitabilne, neurolink i grok u razvoju. Ne mogu da te ubedim u nesto kada si vec odlucio protiv svega. A i mrzi me. Meni je tesla samo profit do sada doneo, nema sta da se zalim. Ljudi su zadovoljni kolima, neki prezadovoljni. Medijske, naucne, drzavne firme koriste spacex da izbace svoje satelite, cena oko 100K sto i nije skupo. Edited November 24, 2023 by mustang
Shan Jan Posted November 24, 2023 Posted November 24, 2023 1 hour ago, Lord Protector said: To je ono što mnogi ne razumeju, Elon umrežava sve svoje kompanije u jedan veliki data pool koji koristi podatke i modele iz jedne firme da bi trenirao AI modele u drugim. To jedan BMW, Mercedes ili Toyota ne mogu da isprate, ograničeni su svojom uskom specilizacijom. Da ne pominjem koliko mu pomaže know-how novih materijala za SpaceX koje koristi u pravljenju novih mehaničkih i elektronskih komponenti za autoindustriju. Sve konvergira ka autonomnim agentima, sve postaje robot. Google ima verovatno mnogo vise podataka nego Elon. Takodje ovaj AI kakav danas postoji tesko da ce zameniti ljude, bar tamo gde se greske skupo placaju. Lepo je kad vidimo sta moze da uradi sa umetnoscu ili koliko dobre odgovore ponekad daje na pitanja ali on, kao sto je ovaj i rekao, samo oponasa ljudsko ponasanje. I to radi na osnovu pravila koje same razvije a kojih ljudi nisu svesni. Pa kad dodje do slucaja u kojem pravila ne vaze i ljudi izginu onda kazemo samo jos malo, naucice... Ja se vise brinem da ce budale, kao sto je krenulo u USA, da guraju taj falican AI tamo gde mu nije mesto, nego da ce to da zameni ljude. Sto se ulaganja tice ne znam... Precenjeni su ali su postali i too big to fail pa ne sumnjam da ce ih prijatelji iz USA establismenta gurati dok god moze.
Sestre Bronte Posted November 24, 2023 Posted November 24, 2023 1 hour ago, Lord Protector said: To je ono što mnogi ne razumeju, Elon umrežava sve svoje kompanije u jedan veliki data pool koji koristi podatke i modele iz jedne firme da bi trenirao AI modele u drugim. To jedan BMW, Mercedes ili Toyota ne mogu da isprate, ograničeni su svojom uskom specilizacijom. Da ne pominjem koliko mu pomaže know-how novih materijala za SpaceX koje koristi u pravljenju novih mehaničkih i elektronskih komponenti za autoindustriju. Sve konvergira ka autonomnim agentima, sve postaje robot. Koliko je novih materijala napravljeno za SpaceX i ko ih je napravio? Na ulazim u priču oko AI, nisam školovao.
Lord Protector Posted November 24, 2023 Posted November 24, 2023 (edited) 17 minutes ago, Braća Strugacki said: Koliko je novih materijala napravljeno za SpaceX i ko ih je napravio? Na ulazim u priču oko AI, nisam školovao. Ima ovde nešto (bajato) o tome https://electrek.co/2018/05/10/tesla-spacex-new-materials/ Edited November 24, 2023 by Lord Protector
Sestre Bronte Posted November 24, 2023 Posted November 24, 2023 ok, znači po pitanju materijala duplo golo, verujem da je ovo ostalo bolje 3
Lord Protector Posted November 24, 2023 Posted November 24, 2023 (edited) 52 minutes ago, Shan Jan said: Google ima verovatno mnogo vise podataka nego Elon. Takodje ovaj AI kakav danas postoji tesko da ce zameniti ljude, bar tamo gde se greske skupo placaju. Lepo je kad vidimo sta moze da uradi sa umetnoscu ili koliko dobre odgovore ponekad daje na pitanja ali on, kao sto je ovaj i rekao, samo oponasa ljudsko ponasanje. I to radi na osnovu pravila koje same razvije a kojih ljudi nisu svesni. Pa kad dodje do slucaja u kojem pravila ne vaze i ljudi izginu onda kazemo samo jos malo, naucice... Ja se vise brinem da ce budale, kao sto je krenulo u USA, da guraju taj falican AI tamo gde mu nije mesto, nego da ce to da zameni ljude. Sto se ulaganja tice ne znam... Precenjeni su ali su postali i too big to fail pa ne sumnjam da ce ih prijatelji iz USA establismenta gurati dok god moze. Nešto se krupno ovih nedelja desilo u OpenAI, ako si pratio dešavanja, pričaju o značajnom proboju prema AGI. Ceo haos oko Altmana i ekipe izgleda ima tu pozadinu. Ako postignu AGI 5. nivo autonomije vozila postaje realan u nekom momentu. edit: Google trenutno kaska https://www.reuters.com/technology/sam-altmans-ouster-openai-was-precipitated-by-letter-board-about-ai-breakthrough-2023-11-22/ Quote Some at OpenAI believe Q* (pronounced Q-Star) could be a breakthrough in the startup's search for what's known as artificial general intelligence (AGI), one of the people told Reuters. OpenAI defines AGI as autonomous systems that surpass humans in most economically valuable tasks. Given vast computing resources, the new model was able to solve certain mathematical problems, the person said on condition of anonymity because the individual was not authorized to speak on behalf of the company. Though only performing math on the level of grade-school students, acing such tests made researchers very optimistic about Q*’s future success, the source said. Edited November 24, 2023 by Lord Protector
Lord Protector Posted November 24, 2023 Posted November 24, 2023 (edited) 27 minutes ago, Braća Strugacki said: ok, znači po pitanju materijala duplo golo, verujem da je ovo ostalo bolje Jel se sećaš koliko su bili kvalitetni SAAB automobili, dok se proizvođač nije u potpunosti fokusirao na gripene? To je ta priča. Edited November 24, 2023 by Lord Protector
Lord Protector Posted November 24, 2023 Posted November 24, 2023 (edited) Quote Musk’s data streams The fuel for AI is data. The new chatbots were being trained on massive amounts of information, such as billions of pages on the internet and other documents. Google and Microsoft, with their search engines and cloud services and access to emails, had huge gushers of data to help train these systems. What could Musk bring to the party? One asset was the Twitter feed, which included more than a trillion tweets posted over the years, five hundred million added each day. It was humanity’s hive mind, the world’s most timely data set of real-life human conversations, news, interests, trends, arguments, and lingo. Plus it was a great training ground for a chatbot to test how real humans react to its responses. The value of this data feed was not something Musk considered when buying Twitter. “It was a side benefit, actually, that I realized only after the purchase,” he says. Twitter had rather loosely permitted other companies to make use of this data stream. In January, Musk convened a series of late-night meetings in his Twitter conference room to work out ways to charge for it. “It’s a monetization opportunity,” he told the engineers. It was also a way to restrict Google and Microsoft from using this data to improve their AI chatbots. There was another data trove that Musk had: the 160 billion frames per day of video that Tesla received and processed from the cameras on its cars. This data was different from the text-based documents that informed chatbots. It was video data of humans navigating in real-world situations. It could help create AI for physical robots, not just text-generating chatbots. The holy grail of artificial general intelligence was building machines that could operate like humans in physical spaces, such as factories and offices and on the surface of Mars, not just wow us with disembodied chatting. Tesla and Twitter together could provide the datasets and the processing capability for both approaches: teaching machines to navigate in physical space and to answer questions in natural language. Edited November 24, 2023 by Lord Protector
gospa buba Posted November 24, 2023 Posted November 24, 2023 44 minutes ago, Braća Strugacki said: ok, znači po pitanju materijala duplo golo, verujem da je ovo ostalo bolje pa to ne mozemo ni da saznamo dok nekad ne ispadne s kamiona. ovako samo novi materijali bazirani na A, B, C, D itd.
Weenie Pooh Posted November 24, 2023 Posted November 24, 2023 2 hours ago, Lord Protector said: To je ono što mnogi ne razumeju, Elon umrežava sve svoje kompanije u jedan veliki data pool koji koristi podatke i modele iz jedne firme da bi trenirao AI modele u drugim. To jedan BMW, Mercedes ili Toyota ne mogu da isprate, ograničeni su svojom uskom specilizacijom. Da ne pominjem koliko mu pomaže know-how novih materijala za SpaceX koje koristi u pravljenju novih mehaničkih i elektronskih komponenti za autoindustriju. Ovo je direktna muskovska propaganda, godinama unazad je slušamo. Bazirana je ni na čemu osim na lepim željama i slepoj veri njegovih najvernijih fanova. Bezbroj je problema preko kojih komotno prelaziš u samo tom jednom pasusu, pobrojaću samo nekoliko: 1. "Data pool" ne znači jedan univerzalni resurs koji se može primenjivati across the board. Podaci iz Tesle su mahom beskorisni za Tviter, a podaci sa Tvitera su mahom beskorisni za SpaceX, a podaci iz SpaceX su mahom beskorisni za Boring Company, itd. "On će to sve da umreži" ne znači ništa. 2. Sve te "usko specijalizovane" kompanije koje "ne mogu da isprate" Teslu su već daleko ispred Tesle po parametru koji se u tvom tekstu navodi kao jedini relevantan, miles per disengagement. Teslin je trenutno samo desetak milja. Waymo (Google) i Cruise (General Motors) beleže desetine hiljada milja. 3. "Knowhow novih materijala" je žvaka za ludaka koju Elon prodaje godinama i godinama. Bušićemo tunele tungstenom koji NASA planira da koristi na asteroidima, a u Cybertruck ćemo ugraditi specijalne raketne pogone! Source je uvek "trust me bro", a followthrough je uvek nepostojeći. 4. "It's like ChatGPT but for cars" je debilni elevator pitch, uopšte me ne iznenađuje da ga je neko prodao Elonu. Ali jezik nije isto što i vožnja. Orijentisanje u prostoru je neuporedivo starije od jezika. Ljudski oko i optički nerv su proizvod ~540 miliona godina evolucije, to nam je dalo spatial reasoning kakav danas imamo, to ne može da se sustigne nikakvim training modelom, ma koliko podataka prikupio. 2 hours ago, Lord Protector said: Sve konvergira ka autonomnim agentima, sve postaje robot. Techno Utopianism takođe počiva na slepoj veri. Ali neka bude da si u pravu, neka sve konvergira ka robotskom singularitetu - to opet ne znači da je Elon Musk taj koji nas vodi u ovu svetlu budućnost. 2 3
Sestre Bronte Posted November 24, 2023 Posted November 24, 2023 25 minutes ago, Lord Protector said: Jel se sećaš koliko su bili kvalitetni SAAB automobili, dok se proizvođač nije u potpunosti fokusirao na gripene? To je ta priča. SAAB je priča iz drugog vremena, a ideja da je BMW ili VW u nekom podređenom položaju po pitanju razvoja je fantastična za sve auto proizvođače, pored njih samih, rade TIER1 firme koje razvijaju, pored ostalog i materijale, broj inženjera koji radi u njima je u stotinama hiljada ideja da neka firma sama može da se nosi sa svim što čini konkurenciju je apsurdna jedina prednost tesle je što nije troma kao stari oem proizvođači, samo to nije baš tako očita prednost, ukoliko nema obećavane revolucije, koja samo što nije, bar 15 godina 4
Lord Protector Posted November 24, 2023 Posted November 24, 2023 (edited) Quote "Data pool" ne znači jedan univerzalni resurs koji se može primenjivati across the board. Podaci iz Tesle su mahom beskorisni za Tviter, a podaci sa Tvitera su mahom beskorisni za SpaceX, a podaci iz SpaceX su mahom beskorisni za Boring Company, itd. "On će to sve da umreži" ne znači ništa. Twitter koristi velike video strimove između ostalog, ne samo tekst. Podaci jesu različiti ali je arhitektura ispod haube ista: LLM transformer model. 2. The power of vertical integration The potential of Tesla bringing battery cell design in-house fits into the company’s long-standing vertical integration approach. Most auto manufacturers take a horizontal integration approach and rely on a myriad of suppliers for both design and production of parts. In contrast, Tesla designs the majority of its parts in-house but outsources the production of most parts. The disadvantage of vertical integration is billions of dollars in capital expenditures related to factories and production equipment, along with the high cost of recruiting and maintaining in-house technical design and testing talent. The most significant advantage is that product design cycle times decrease, which quickens the pace of innovation, which is Tesla’s central competitive moat. Additionally, the end product typically yields better performance and reliability compared to products built with horizontal integration. Vehicle range is one example of a performance advantage Tesla has gained through vertical integration. Teslas deliver greater range with similar-sized batteries compared to other carmakers. Independent Tesla analyst Matt Joyce has been tracking this dynamic since 2019 and has created a core efficiency rating to compare EV batteries. The metric factors in battery size (in kWh) along with the range and weight of the vehicle. A lower number represents greater vehicle efficiency, a combination of the battery pack, electric motor, and wind drag coefficient. In the table below (updated in June 2020), the average core efficiency for Tesla is 5.5, compared to 7.8 for the 15 other EVs available in the US market. This suggests Tesla delivers 42% more efficiency per kWh compared to the average EV model. Quote "It's like ChatGPT but for cars" je debilni elevator pitch, uopšte me ne iznenađuje da ga je neko prodao Elonu. Ali jezik nije isto što i vožnja. Orijentisanje u prostoru je neuporedivo starije od jezika. Ljudski oko i optički nerv su proizvod ~540 miliona godina evolucije, to nam je dalo spatial reasoning kakav danas imamo, to ne može da se sustigne nikakvim training modelom, ma koliko podataka prikupio. I jezik i vožnja, i orijentacija u prostoru su algoritamski, računarski problem. U osnovi imaš određeni matematički ili logički model koji preslikavaš na neuronsku mrežu. Kompleksnost je različita ali logičko-matematičko-algoritamska osnova je ista (Curry-Howard-ova korespodencija, BHK interpretacija). Semantika NLP je još sedamdesetih radom logičara Richarda Montague prebačena u polje lambda računa i logike. Vožnja je klasičan decision problem, sa time se AI bakće već 50 godina. Orijentisanje u prostoru je u osnovi problem kombinatorike i optimizacije. Sa deep learning-om on postaje proces optimizacije u okviru neuronske mreže. U suštini, neuronska optimizacija je ubrzana mašinska evolucija, vršiš virtuelnu razmenu velike količine podataka i paralelnog procesorskog vremena za linearno evolutivno vreme. Npr u oblasti veštačke inteligencije genetski algoritam (GA) je pretraživačka heuristika koja oponaša proces prirodne selekcije. Itd... Naravno, daleko je to od AGI i živog čoveka. To je nedostižno, ljudska inteligencija konvergira ka beskonačnosti, transhumanističke mantre su pogrešne, Gedel je još dvadesetih godina dokazao da su svi ovi (naizgled) savršeni modeli i teorije inherentno puni logičkih paradoksa. Ingenioznost i nove ideje ne mogu se obuhvatiti nekom matematičko-logičkom teorijom. Intuicija je dar posebne vrste, van domašaja formalne logike i algoritama. Ono što transhumanizam pokušava je stvaranje mitskog Golema, preteče robota, bića od blata (silikona, osnovnog sastojka procesora ). Ali Golem nije čovek od krvi i mesa, on je nemi rob, sredstvo. Nikad neće biti na istom nivou kao čovek. Ali neki zadaci, koji su repetitive prirode će moći da se automatizuju. Vožnja, snalaženje u prostoru, semantika jezika. Nove ideje i intuicija ne. Edited November 24, 2023 by Lord Protector
Weenie Pooh Posted November 24, 2023 Posted November 24, 2023 (edited) 35 minutes ago, Lord Protector said: Core Efficiency Pa tekst koji si sam okačio bavi se miles per disengagement, dakle distancom vožnje bez uočene greške. Tesla je tu mizerno loša, što bi trebalo da govori nešto o sposobnostima njihove self-driving tehnologije. Ali neee, hajde bolje da gledamo odnos cene i core efficiency, što je indeks jačine baterije, dometa, i težine vozila! To će nam reći da li self-driving radi ili ne radi Vertikalna integracija je isto neviđena prednost. Kao, mi autsorsujemo samo proizvodnju, ali dizajn radimo in-house, zato smo ovako inovativni dizajneri Koji inovativni dizajn, alo? Gde je inovacija, u onom nesrećnom kockastom Cybertrucku? Šta Tesla radi što ne radi konkurencija brže, bolje, i za manje para? Navodi se da im je inovacija donela bolji range za baterije istog kapaciteta, što je notorna glupost. Teslin range je ekstremno nepouzdan, evo šta je Forbes imao o tome da kaže letos: Quote Tesla programmed its dashboards to exaggerate how far the vehicles can drive on a single charge, and canceled service appointments for customers who noticed their cars' range was shorter than expected, a report by Reuters found Thursday — the latest Tesla feature to face scrutiny. Dakle, jedina prava inovacija koju tesla nudi je u besramnom laganju - tu veštinu su inovativno uzdigli do rekordnih nivoa. 35 minutes ago, Lord Protector said: Ali neki zadaci, koji su repetativne prirode će moći da se automatizuju. Vožnja, snalaženje u prostoru, semantika jezika. Nove ideje i intuicija ne. Opet svrstavaš spatial orientation u isti koš sa semantikom, za šta zaista nema osnova. Ljudska reakcija na vizuelni imput je u ogromnoj meri intuitivna - rizik se procenjuje bez razmišljanja, reakcije se izvode heuristički, kognicija tu leti kroz prozor. Dakle po svemu sudeći nije u pitanju "decision problem", algoritamski rešiv. Sasvim je moguće da FSD nikad nećemo videti, tj. da će kompanije koje jure taj sveti gral uskoro odustati. Reuters: Quote The undelivered promise of true autonomy has raised the stakes for the AV industry. "If these companies don't succeed over the next two years, they're not going to exist anymore," said Mike Wagner, CEO of Edge Case Research, which helps AV companies assess, manage and insure risk. "It's a case of put up or shut up at this point." Naravno, razlika je u tome što BMW, Mercedes, ili Toyota neće propasti ako se ispostavi da su uzalud jurili autonomous driving. Tesla hoće, jer Tesla ništa drugo nema. Ima samo apsurdno prenaduvanu tržišnu vrednost, baziranu na fantomu robotske vožnje. Bez toga joj sleduje kolaps. Edited November 24, 2023 by Weenie Pooh 2 1
Lord Protector Posted November 24, 2023 Posted November 24, 2023 (edited) 1 hour ago, Weenie Pooh said: Opet svrstavaš spatial orientation u isti koš sa semantikom, za šta zaista nema osnova. Ljudska reakcija na vizuelni imput je u ogromnoj meri intuitivna - rizik se procenjuje bez razmišljanja, reakcije se izvode heuristički, kognicija tu leti kroz prozor. Dakle po svemu sudeći nije u pitanju "decision problem", algoritamski rešiv. To je SLAM problem Quote Simultaneous localization and mapping (SLAM) is the computational problem of constructing or updating a map of an unknown environment while simultaneously keeping track of an agent's location within it. While this initially appears to be a chicken or the egg problem, there are several algorithms known to solve it in, at least approximately, tractable time for certain environments. Popular approximate solution methods include the particle filter, extended Kalman filter, covariance intersection, and GraphSLAM. SLAM algorithms are based on concepts in computational geometry and computer vision, and are used in robot navigation, robotic mapping and odometry for virtual reality or augmented reality. https://en.wikipedia.org/wiki/Simultaneous_localization_and_mapping Edited November 24, 2023 by Lord Protector
Lord Protector Posted November 24, 2023 Posted November 24, 2023 (edited) malo da se igramo sa Anthropic Claude 2: give me some slam problem, than convert it to montague grammar specification, than convert it to lambda calculus Here is a slam problem example converted into a Montague grammar specification and then into lambda calculus: Slam Problem Robot A starts at location (2,2). It moves right 3 units and up 2 units. It senses landmark L at location (7,5). Montague Grammar S -> NP VP VP -> IV | TV NP NP -> PN | Det N | Pro N -> 'robot' | 'location' | 'landmark' IV -> 'moves' | 'senses' TV -> 'starts at' | 'is at' Det -> 'a' | 'the' Pro -> 'it' PN -> 'A' | 'L' This context-free grammar decomposes sentences into noun phrases (NP) and verb phrases (VP), with production rules defining terminal vocabulary. Lambda Calculus A = λx.robot(A) start = λxyr.startsAt(r,x,y) move = λdxλdyλr.moved(r,dx,dy) L = λxy.landmark(L,x,y) sense = λlr.sensed(l,loc(r)) start(2)(2)(A) move(3)(2)(A) sense(L(7)(5))(A) Where: start, move, sense are semantic predicates r, l, x, y are arguments The lambda terms compositionally build up meaning. Let me know if you need any clarification or have additional examples to walk through! Here is a concrete example of how we could formulate parts of the previous slam problem with a deep learning approach: Problem Robot A starts at (2,2) on a 10x10 grid map. It moves 3 units east and 2 units north. It then senses the location of landmark L at (7,5). Data We will train a model on a dataset of examples mapping robot sensor data to landmark locations. Examples have the form: { "start_pos": [2, 2], "movement": [+3, +2], "sensor_data": [0.1, 0.2, ..., 0.5], "landmark_pos": [7, 5] } Model Our model is a neural network with the following architecture: Input layer taking in start position, movement vector, and sensor readings 3 Hidden layers of 64 nodes each Output layer with 2 nodes predicting the landmark x,y coordinates Training We train this model on 10000 examples showing different start positions, movements, sensor data, and landmark locations, minimizing the mean squared error loss between predictions and actual landmark coordinates. Inference Given new input data like that shown above, our trained model takes the start position, movement, and sensor data, and predicts the landmark location (x,y) coordinates. We could expand on this with additional network outputs to predict a more complete map with multiple landmarks, or replace the hardcoded movement vector with additional sensor inputs to have the model infer movements as well. But hopefully this gives a concrete idea of one way to formulate parts of the slam problem with a data-driven deep learning approach! Edited November 24, 2023 by Lord Protector
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now