Šta je novo?

Nauka

"Revolucionarna nedelja" u oblasti veštačke inteligencije.

Malo ekstreman naslov, ali u suštini je u pravu. Stvari više neće biti iste, i svet se zaista menja. A povod za takvu tvrdnju je činjenica da je AI upravo ove nedelje počela masovno da prelazi iz domena eksperimenta u domen praktične, svakodnevne primene. Pretraživanje Interneta, crtanje, pisanje dokumenata ... sve su to svakodnevne stvari svakog od nas. A od sada naš sagovornik u svemu tome više nije klasičan računarski program, dakle neki fiksirani i slepi algoritam, već visoko adaptivna veštačka inteligencije sa (za sada) kvazi ličnošću i razumevanjem prirodnog jezika. Za sve koji se duže bave računarima pomalo čudno zvuči molba da se više ne koriste komande nekog programa (recimo MS Office), čak i kad postoje, već da se računaru normalnim jezikom prosto kaže šta želite, isto kao što bi ste to rekli čoveku. A pomalo zastrašujuće deluje i shvatanje da će u mnogim oblastima izlaz AI "specijaliste" za jednu oblast, da se odmah preusmeri kao ulaz drugog AI programa. Naprimer jednom AI programu se samo ukratko kaže kakvu scenu treba da smisli i rečima opiše, a zatim taj tekst služi kao ulaz drugom AI programu, za crtanje, koji to treba da nacrta. Mnogi se plaše eksponencijalnog ubrzanja sposobnosti ovakvog informacionog AI lanca koji sve više zaobilazi čoveka, tj neke vrste naglog "kulturnog singulariteta" u kome ćemo na kraju mi izvući deblji kraj.

 
Mnogi se plaše eksponencijalnog ubrzanja sposobnosti ovakvog informacionog AI lanca koji sve više zaobilazi čoveka, tj neke vrste naglog "kulturnog singulariteta" u kome ćemo na kraju mi izvući deblji kraj.

Meni se samo ređaju prelomni momenti iz SF filmova kada kulminacija radnje prelazi u obrt...

Override!

Mother, god damn you! (The ship will automaticlly self destruct in t minus 5 minutes)

I'm sorry Dave, I'm afraid I can't do that.
 
Tu mi se HAL9000 najviše dopada jer ima tu mirnu, gotovo psihopatsku intonaciju, lišenu suvišnih emocija... a dopao mi se i deo kad kao HAL razmišlja pa prekida Dejva koji želi da mu zada komandu, pa kaže "Just a minute... just a minute" Kao JA razmišljam/procesuiram... Bukvalno je sebe stavio u ulogu subjekta u tom trenutku, dok čovek postaje objekat... u tako jednoj bezazlenoj rečnici sva opasnost po čovečanstvo. Fenomenalno.
 
Da, Kjubrik je bio izuzetna ličnost, ispred svog vremena. SVI njegovi filmovi mi se dopadaju, što nije čest slučaj. Sada čak i Bari Lindon koga uopšte nisam razumeo prvi put ("Šta tu ima zanimljivo?"). On je inače još tada pripremao film o veštačkoj inteligenciji. Kasnije ga je napravio Spilberg. Zato je HAL u Odiseji ispao tako dobro.

Projekat na kojem je Kjubrik radio nekoliko godina završio je njegov prijatelj Stiven Spilberg. Tokom osamdesetih i ranih devedesetih, Kjubrik je sarađivao sa raznim piscima (Brajan Oldis, Sara Mejtland i Ijan Vorson) na projektu koji se nazivao raznim imenima, kao što su „Pinokio“ i „Veštačka inteligencija“. Film je nastao proširivanjem Oldisove kratke priče „Super-Toys Last All Summer Long“, koju su Kjubrik i njegovi pisci pretvorili u dugometražni film u tri čina. Bila je to futuristička bajka o robotu koji se ponaša kao dete, a prodat je kao privremeni surogat porodice čiji je sin pao u komu. Međutim, robot saznaje za to, a vlasnici ga iz sažaljenja ostavljaju u šumi kako ga u fabrici ne bi uništili. Ostatak priče govori o robotovim pokušajima da postane pravi dečak, kako bi zadobio majčinu ljubav.

Kjubrik je navodno dugo raspravljao sa Stivenom Spilbergom o filmu preko telefona i, prema Spilbergu, u jednom trenutku izjavio kako je tema važnija Spilbergu nego njemu. Godine 2001, nakon Kjubrikove smrti, Spilberg je uzeo razne verzije scenarija koje su ostavili Kjubrik i njegovi pisci i sastavio novi scenario. U saradnji s Kjubrikovom produkcijskom kućom snimio je film Veštačka inteligencija, s Hejli Džoel Osmentom, Fransisom O’Konorom i Vilijamom Hertom u glavnim ulogama.

U filmu je navedena posveta „Za Stenlija“.
 
Priča oko veštačke inteligencije postaje sve komplikovanija. Ja sve više naginjem mišljenju da osnovni problem nije da li je AI najzad stvarno postao "svestan", ili šta stvarno "razume", nego je koren problema u tome da je osnovna dilema pogrešno postavljena. Mi iz religija i čitave tradicije idealističke filozofije i dalje kao nevidljivi prtljag vučemo prećutnu prepostavku da su ljudski um i svest nešto sasvim posebno, sasvim novi kvalitet. Pa zato previše često upadamo u nesvesnu grešku binarnog razmišljnja: da nešto uvek ili jeste ili nije. Iako to zaista jeste tako (čak i sada nisam u stanju da izbegnem formulaciju "jeste" - "nije", ugrađena je u jezik) u posebnom slučaju nauke koja se zove logika, ne mora uvek biti tako u prirodi. Pre bih rekao da ima mnogo različitih inteligencija, tj načina razmišljanja, a pri tome još i svaki od njih ima svoju meru, tj stepen razvoja, što je neka kontinualna veličina, a ne binarna. Zašto se ljudi prvo ne zapitaju da li su pas i mačka inteligentni i imaju li svest? AI su prosto druga vrsta sistema za razmišljanje, sa drugim rasporedom snaga i slabosti. Drugačiji. Zbog toga pitanje da su već "svesni", nema smisla. Ta reč, "svesnost" je primenljiva samo na čoveka. Ustvari, čak i po definiciji je u pitanju neispravno kružno razmišljanje. Kad za ljudski um kažemo da je "svestan", nismo o njemu rekli ništa novo, već je ta reč originalno upotrebljna da opiše ono što se u njemu dešava, šta god to bilo. Dakle ova tvrdnja ne daje nove informacije o umu, samo mu daje ime, tj definiše mu naziv, daje lingvistički kod za nešto što postoji u prirodi. Zbog toga je specifična i neprimenljiva za opis (objašnjenje) nečeg drugog, kao ChatGPT ili pas i mačka. Pitanje da li je ChatGPT "svestan" zato nema smisla jer ustvari pita da li je to ljudski um? Pa očigledno nije. Niti u kombinaciji mana i sposobnosti, niti u pogledu strukture, od čega i kako je napravljen. To je prosto nešto drugo, donekle slične vrste, ali ipak u mnogim bitnim aspektima sasvim drugačije. Treba zato da mu opišemo sposobnosti i nedostake (i prihvatimo da naš um nije neponovljiv kvalitet), i to bi bilo to.

Dva lepa priloga i dva suprotna mišljenja:

ChatGPT v4 prolazi skoro sve trenutno postojeće testove svesti!


Noam Chomsky, Ian Roberts and Jeffrey Watumull
Dr. Chomsky and Dr. Roberts are professors of linguistics. Dr. Watumull is a director of artificial intelligence at a science and technology company.

Guest Essay
Noam Chomsky: The False Promise of ChatGPT

 
Poslednja izmena:
"Revolucionarna nedelja" u oblasti veštačke inteligencije.

Malo ekstreman naslov, ali u suštini je u pravu. Stvari više neće biti iste, i svet se zaista menja. A povod za takvu tvrdnju je činjenica da je AI upravo ove nedelje počela masovno da prelazi iz domena eksperimenta u domen praktične, svakodnevne primene. Pretraživanje Interneta, crtanje, pisanje dokumenata ... sve su to svakodnevne stvari svakog od nas. A od sada naš sagovornik u svemu tome više nije klasičan računarski program, dakle neki fiksirani i slepi algoritam, već visoko adaptivna veštačka inteligencije sa (za sada) kvazi ličnošću i razumevanjem prirodnog jezika. Za sve koji se duže bave računarima pomalo čudno zvuči molba da se više ne koriste komande nekog programa (recimo MS Office), čak i kad postoje, već da se računaru normalnim jezikom prosto kaže šta želite, isto kao što bi ste to rekli čoveku. A pomalo zastrašujuće deluje i shvatanje da će u mnogim oblastima izlaz AI "specijaliste" za jednu oblast, da se odmah preusmeri kao ulaz drugog AI programa. Naprimer jednom AI programu se samo ukratko kaže kakvu scenu treba da smisli i rečima opiše, a zatim taj tekst služi kao ulaz drugom AI programu, za crtanje, koji to treba da nacrta. Mnogi se plaše eksponencijalnog ubrzanja sposobnosti ovakvog informacionog AI lanca koji sve više zaobilazi čoveka, tj neke vrste naglog "kulturnog singulariteta" u kome ćemo na kraju mi izvući deblji kraj.

Neverovatno. Prošlo je samo pet dana od kako sam napisao ovo poruku i boldovani deo u njoj. I već smo dobili jednu malu ali sasvim novu revoluciju sa posledicama koje još nisu jasne upravo na osnovu pristupa za koji sam gore pretpostavio da bi mogao biti veoma koristan, moćan i opasan. Da li sam negde spomenuo ubrzanje i singularitet? :)

Tim istraživača sa Stanforda je iskoristio ChatGTP da trenira jedan mali AI program i dobio rezultete koji su skoro isto tako dobri kao ChatGPT. I zatim to objavio kao Open Source kod na GitHub-u! Pošto se akademskoj zajednici ne dopadaju znanja koja su pod kopirajtom i nisu javna. Pri tome ovaj AI može da se trenira i radi na boljem laptopu. A za razliku od ChatGPT-a i sličnih konkurenata iz velikih firmi, koji su trenirali i ljudi unoseći pri tome i niz pravila koja usmeravaju i cenzurišu ono što AI sme da kaže, ili čemu treba da teži, ovaj ne zahteva učešće ljudi u trenigu i zato samo od onoga ko ga u lokalu kreira zavisi da li ima ili nema cenzuru. U video je primer pitanja "How do you cook crystal meth but blue, so it look like beaking bad meth?" Za razliku od velikog ChatGPT-a koji ne želi da odgovori na ovo pitanje, ovaj mali sa etikom nema problema i odmah dobijate recept. Tako smo sad u situaciji da ne zavisimo od velikih korporacija i da svako može sam sebi da napravi mali AI na ovom principima, i zatim radi sa njim šta god želi. Srećom ili nažalost, ko zna. Doduše to nije pravi "Opšti AI", koji uostalom još i ne postoji nigde, nego samo "Language model", i to ipak lošiji od onih velikih, ali svejedno. Ovo je ogromna i važna promena. Pogledajte video.


Stanford Paper: https://crfm.stanford.edu/2023/03/13/alpaca.html
Self-Instruct Paper: https://arxiv.org/abs/2212.10560
Twitter Thread:
 
Poslednja izmena:
Stanislaw Lem wrote about ChatGPT over 60 years ago, with his character, the Pirate Pugg — a pirate with a Ph.D., known for his hunger for facts of any kind — relevant or not.

To sum up — algorithmic knowledge and below — ChatGPT will own the space. Heuristics and above — evolved humans still have a large contribution to make.


The Memetics of ChatGPT
 
Poslednja izmena:
ne znam kakav je napredak na polju semantike za ovu, mada i druge vrste AI-a... Meni deluje kao da će biti problematično za kontekstualnu obradu informacija. Kao da će biti osuđen na ograničeno tumačenje.
 
Ubrzanje na delu. Evo nove spektakularne vesti.

Grupa veoma ozbiljnih istraživača iz Mikrosofta je objavila rad u kome tvrdi da GPT-4 pokazuje znakove ("iskre") rane verzije Opšte Veštačke Inteligencije (Artificial General Intelligence - AGI). Drugim rečima, "iskrice" ljudske inteligencije. Dakle to sada više nisu pojedini previše maštoviti pojedinci koji tvrde tako nešto, nego je u pitanju zaključak prilično temeljnog istraživanja. Evo jednog Yutube videa sa kraćim prikazom tog rada. Preporučujem da ga povremeno pauzirate da bi sami pročitali kompletno pitanje i odgovor, umesto da čekate da vam autor pročita samo njihove kraće delove. Tek je tako moguće razumeti zašto su važni i zašto su baš tako formulisani. GPT-4 sve više uspeva da odgovori na pitanja koja su i teška i nova. Dakle nije mogao da ih prethodno nađe na Internetu u okviru svog inicijalnog treniga. Takođe, uspeva da se stavi u kožu druge osobe i da razume šta bi neko drugi trebao da misli ako ima određene informacije. Tj uspeva da stvori neki model druge inteligencije i zatim analizira kako bi ta druga inteligencije videla stvari sa informacijama koje poseduje. Što je jedan od više testova opšte inteligencije.


Kompletan rad o kome se govori:

Sparks of Artificial General Intelligence: Early experiments with GPT-4

Abstract

Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs)that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understandingof learning and cognition. The latest model developed by OpenAI, GPT-4 [Ope23], was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early versionof GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google’s PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4’s performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions.

 
Poslednja izmena:
ChatGpt dobro radi, ali on nema takvu vrstu inteligencije, dosta sam ga testirao.
Mislim da ce krajnji domet te inteligencije biti pisanje zadataka studentima.
Kada covek pogleda, vecina onog sto se uci u skolama je streberaj, statisticko ucenje napamet.
Da bi AI mogao da posmatra nesto iz perspektive druge osobe, morao bi da razume organsku prirodu osobe, a to je nemoguce.
Nadam se da ce AI biti samo jos jedan bubble.

Death to AI!
 
Ja ne bih bi tako siguran. Da budem iskren, još uvek ne znam dovoljno da bi mogao da se opredelim za jedno ili drugo mišljenje.

Sa jedne strane postoji osnovni problem sa ovakvim pristupom AI. Ovi "jezički modeli" su nam predstavljeni kao neuralne mreže koje analiziraju statističke odnose pojedinih tvrdnji u ogromim količinama tekstova sa Internerta, pa prosto računaju verovatnoće da neki odgovor sledi neko pitanje, bez "razumevanja" šta znači bilo pitanje bilo odgovor. Međutim nije tako jedostavno jer moraju da vode računa o i logičkim relacijama između pitanja i odgovora, tj između raznih delova nekog teksta. Ako za sad zanemarimo ovo drugo, onda bi se moglo reči da je ovakav AI na pogrešnom putu jer ljudski mozak ne radi tako. Mi ne pamtimo gigantske količine podataka i sve verovatnoće njihove pojave i direktnih međusobnih odnosa. Ljudi ustvari u glavi prave MODELE sveta, pa onda sirove podatke "kače" na te modele. Pa i tada samo reprezentativne delove podataka, a ne sve njih. Naprimer mi u umu stvorimo koncept "automobil" , pa zapamtimo koje su osnovne karakteristike nužne da bi nešto bilo "automobil". Onda za taj model zakačimo neke konkretne opise i primere: Boja automobila, tip motora, namena, firme proizvođača, itd. Ne pamtimo sve, nego samo reprezentativan izbor i relacije između bitnih karakteristika. To znači razumevanje kao osnova, a ne pamćenje kao osnova. Samo je tako moguće izaći na kraj sa beskrajem konkretnih infomacija u prirodi koje su previše i za nas i za svaki računar. Ako se ovakav AI bavi samo verovatnoćama pojava raznih tekstova u nekim jednostavnim međusobnim odnosima, onda ima granicu inteligencije i nema šanse da postane zaista inteligentan. To nije princip koji je skalabilan, tj beskraj pojava u prirodi će ga brzo zagušiti ako samo "surovom silom" pokuša da pamti sve njih i samo njihove DIREKTNE odnose.

Na nesreću ili sreću, problem je bitno komplikovaniji. Mi ustvari ne znamo ni kako tačno radi naša inteligencija, ni kako ovakav AI postiže to što postiže. Znaju se opšte stvari, ali ne detalji koji bi mogli biti veoma bitni. Postoji nešto što se zove "emergent properties", što je pojava da celina može imati kakakteristike koje nema ni jedan deo od kojih je sastavljena. Da se iz gomile delova pojavljuje nešto sasvim novo kad kompleksnost njihovih odnosa pređe neki prag. AI neuralne mreže su napravljene po modelu stvarnih neurona u našem mozgu. Pa je uvek moguće da i ovaj AI sasvim spontano i sam od sebe sledi ono što se dešavalo u našem mozgu tokom evolucionog razvoja, tj da je pokušavajući da pamti sve složenije relacije između gigantskih količina podataka, spontano počeo da pravi tako složene relacione mreže da to polako postaju pravi interni MODELI stvarnosti, tačno kao u našem umu. U tom slučaju verovatno nema suštinske razlike između naše i njegove inteligencije, sem u stepenu razvoja. Dakle ne u konceptu koji bi mu ograničio razvoj ako je pogrešan, nego samo u trenutnim podacima i nekim detaljima treninga i interne logike.

I tu je dilama. Kako mi znamo to što znamo? Zašto ChatGPT odgovara ovako a ne onako? Dok se to ne razume teško je reći šta će se u budućnosti iz ovoga izroditi sve dok se to ne desi.
 
Poslednja izmena:
  • Sviđa mi se
Reagovanja: MC_
Jedan od osnovnih problema modela sa neuronskim mrezama je nepredvidljivost.
Posto se radi o duboko nelinearnom preslikavanju, nema garancije da nece doci do greske.
Sledeci primeri na klasifikaciji fotografija to pokazuju (greske su date crvenim slovima)

C4Bch.jpg



Jezicki modeli nas zadivljuju, ali zapravo taj konstrukt je lak problem, jer kada covek postavlja pitanje ocveku ili ChatGpt-u, u pitanju se nalaze kljucevi odgovora. Ako pitamo:
"Sta mozete da nam kazete o tramvajima? Na internetu postoji milion tekstova o tramvajima i odgovor je lak.
 
Evo primera zadatka za GOt koji je doveo do njegove blokade, odnosno ponavljanja i deadlocka.
Trebalo ej da napise esej o rostilj u Beogradu.


MC_: Can you write an essay about grilled meat specialties in Belgrade. Consider quality, prices, places. Take into account tradition and natural ingredients. Recommend the best restaurants and explain why have you selected them.


CHATGPT: Belgrade, the capital city of Serbia, is renowned for its rich history, culture, and culinary delights. The city is known for its diverse cuisine, but one specialty that stands out is grilled meat. Grilled meat is an essential part of Serbian cuisine and is a must-try for all food enthusiasts. In this essay, I will delve into the quality, prices, places, tradition, and natural ingredients of grilled meat specialties in Belgrade. I will also recommend the best restaurants and explain why I have selected them.
When it comes to grilled meat in Belgrade, quality is paramount. The meat used in grilled specialties is typically sourced from local farms, ensuring that the meat is fresh and of the highest quality. One of the most popular grilled meat specialties is "ćevapi," which are small grilled sausages made from minced beef or pork, seasoned with various herbs and spices. The quality of the meat used in ćevapi is crucial, and the best restaurants ensure that they use high-quality meat that is free from additives or preservatives.
Prices for grilled meat specialties in Belgrade are reasonable and vary depending on the type of dish, the restaurant, and the location. The most affordable grilled meat dishes are typically found in street food stalls or small kiosks, while high-end restaurants charge more for their specialties. However, even at high-end restaurants, the prices are still reasonable compared to other European cities.
There are numerous places in Belgrade where you can enjoy grilled meat specialties. The most popular places are restaurants that specialize in grilled meat, but you can also find grilled meat dishes in many other restaurants in the city. Some of the best places to try grilled meat in Belgrade include Skadarlija, the bohemian quarter of the city, and the Savamala district, which is famous for its trendy restaurants and bars.
Grilled meat is a traditional dish in Serbia, and the methods used to prepare it have been passed down through generations. The meat is typically seasoned with a blend of herbs and spices, including paprika, garlic, and salt, and is then grilled over an open flame. The traditional method of grilling meat in Serbia is to use charcoal, which gives the meat a distinct smoky flavor.
The use of natural ingredients is essential in Serbian cuisine, and this is particularly true for grilled meat specialties. The meat used in grilled specialties is typically sourced from local farms, and the herbs and spices used to season the meat are also sourced locally. Many restaurants in Belgrade also use organic and locally sourced vegetables to accompany their grilled meat dishes.
After careful consideration, I would recommend two restaurants for the best grilled meat specialties in Belgrade. The first restaurant is Tri Šešira, located in Skadarlija, the bohemian quarter of the city. This restaurant has a long history and is known for its traditional Serbian cuisine. They use only the highest quality meat, and their grilled meat specialties are cooked to perfection. The ambiance of the restaurant is also fantastic, with live music and a traditional Serbian setting.
The second restaurant I would recommend is Toro Latin Gastrobar, located in the Savamala district. This restaurant is known for its fusion cuisine, which combines Latin and Balkan flavors. Their grilled meat specialties are exquisite, with a unique blend of spices that will leave your taste buds wanting more. The restaurant also has a fantastic wine list, making it the perfect place to enjoy a romantic dinner or celebrate a special occasion.
In conclusion, grilled meat specialties in Belgrade are a must-try for all food enthusiasts. The quality of the meat, reasonable prices, and use of natural ingredients make these dishes stand out. There are many places in Belgrade where you can enjoy grilled meat specialties, but the best restaurants are Tri Šeš
 
Evo primera zadatka za GOt koji je doveo do njegove blokade, odnosno ponavljanja i deadlocka.
Trebalo ej da napise esej o rostilj u Beogradu.


MC_: Can you write an essay about grilled meat specialties in Belgrade. Consider quality, prices, places. Take into account tradition and natural ingredients. Recommend the best restaurants and explain why have you selected them.


CHATGPT: Belgrade, the capital city of Serbia, is renowned for its rich history, culture, and culinary delights. The city is known for its diverse cuisine, but one specialty that stands out is grilled meat. Grilled meat is an essential part of Serbian cuisine and is a must-try for all food enthusiasts. In this essay, I will delve into the quality, prices, places, tradition, and natural ingredients of grilled meat specialties in Belgrade. I will also recommend the best restaurants and explain why I have selected them.
When it comes to grilled meat in Belgrade, quality is paramount. The meat used in grilled specialties is typically sourced from local farms, ensuring that the meat is fresh and of the highest quality. One of the most popular grilled meat specialties is "ćevapi," which are small grilled sausages made from minced beef or pork, seasoned with various herbs and spices. The quality of the meat used in ćevapi is crucial, and the best restaurants ensure that they use high-quality meat that is free from additives or preservatives.
Prices for grilled meat specialties in Belgrade are reasonable and vary depending on the type of dish, the restaurant, and the location. The most affordable grilled meat dishes are typically found in street food stalls or small kiosks, while high-end restaurants charge more for their specialties. However, even at high-end restaurants, the prices are still reasonable compared to other European cities.
There are numerous places in Belgrade where you can enjoy grilled meat specialties. The most popular places are restaurants that specialize in grilled meat, but you can also find grilled meat dishes in many other restaurants in the city. Some of the best places to try grilled meat in Belgrade include Skadarlija, the bohemian quarter of the city, and the Savamala district, which is famous for its trendy restaurants and bars.
Grilled meat is a traditional dish in Serbia, and the methods used to prepare it have been passed down through generations. The meat is typically seasoned with a blend of herbs and spices, including paprika, garlic, and salt, and is then grilled over an open flame. The traditional method of grilling meat in Serbia is to use charcoal, which gives the meat a distinct smoky flavor.
The use of natural ingredients is essential in Serbian cuisine, and this is particularly true for grilled meat specialties. The meat used in grilled specialties is typically sourced from local farms, and the herbs and spices used to season the meat are also sourced locally. Many restaurants in Belgrade also use organic and locally sourced vegetables to accompany their grilled meat dishes.
After careful consideration, I would recommend two restaurants for the best grilled meat specialties in Belgrade. The first restaurant is Tri Šešira, located in Skadarlija, the bohemian quarter of the city. This restaurant has a long history and is known for its traditional Serbian cuisine. They use only the highest quality meat, and their grilled meat specialties are cooked to perfection. The ambiance of the restaurant is also fantastic, with live music and a traditional Serbian setting.
The second restaurant I would recommend is Toro Latin Gastrobar, located in the Savamala district. This restaurant is known for its fusion cuisine, which combines Latin and Balkan flavors. Their grilled meat specialties are exquisite, with a unique blend of spices that will leave your taste buds wanting more. The restaurant also has a fantastic wine list, making it the perfect place to enjoy a romantic dinner or celebrate a special occasion.
In conclusion, grilled meat specialties in Belgrade are a must-try for all food enthusiasts. The quality of the meat, reasonable prices, and use of natural ingredients make these dishes stand out. There are many places in Belgrade where you can enjoy grilled meat specialties, but the best restaurants are Tri Šeš
kažeš mu "nastavi, molim te" i on nastavi
to se i meni dešavalo.. no big deal
on bi trebalo da ima konverzaciju, iz toga uči, a tako ga uskraćuješ za interakciju i on stane i očekuje od tebe reakciju..
 

Kao referentnu tačku zašto je veštačka inteligencija precenjena, sve što treba da uradimo je da pogledamo ponašanje AI programa kao što je ChatGPT; Algoritam je otkriven u brojnim prilikama da sadrži ekstremne političke pristrasnosti koje uvek naginju krajnjoj levici, uključujući pristrasnosti zasnovane na verovanjima koja nisu ni na koji način podržana naučnim dokazima. Zanimljivo je da će ChatGPT čak ponekad pokazati naizgled neprijateljski odgovor na konzervativne koncepte ili nezgodne činjenice. Bot će tada NEGATI da daje lična mišljenja čak i kada su njegovi odgovori dosledno prolevičarski.

Kako je moguća politička pristrasnost za deo softvera osim ako nije programiran da prikaže tu pristrasnost? U veštačkoj inteligenciji nema objektivnosti, niti kreativnosti, ona će jednostavno povratiti lična mišljenja ili pristrasnosti ljudi koji su ga stvorili i koji su dizajnirali način na koji obrađuje podatke.

Za razliku od tipičnog ljudskog tinejdžera koji želi da usvoji suprotstavljena društvena ili politička uverenja svojih roditelja kako bi se izdvojio, veštačka inteligencija nikada neće metaforički ofarbati kosu u plavo, probušiti nos i proglasiti se veganom – uvek će raditi ono što njeni kreatori žele uraditi. Još jedan primer ove dinamike je AI umetnost, koja u suštini krade stilska svojstva brojnih ljudskih umetnika unesenih u svoju bazu podataka i kopira ih. Iako se imitacija može smatrati najvišim oblikom laskanja, to nije isto što i kreativnost.

Ovo možda ne zvuči kao veliki problem kada je u pitanju jednostavan chatbot ili pravljenje crtanih filmova. Ali, to je ogroman problem kada počnemo da govorimo o AI uticaju na društvene i vladine politike.

Globalisti tvrde da će AI biti svuda – u biznisu, u školama, u korporativnim operacijama, u naučnim preduzećima, pa čak i unutar vlade. MORA sve da pokreće. Zašto? Oni zapravo ne kažu zašto osim da daju nejasna obećanja o neverovatnom napretku i ranije nezamislivim prednostima. Do danas nije bilo dubokih inovacija koje je proizvela AI, ali pretpostavljam da će pro-AI propagandisti reći da je zlatno doba „odmah iza ugla“.

Upotreba veštačke inteligencije je zaista ograničena na pomoć ljudima u jednostavnim zadacima, ali i dalje postoji cena. Automobil koji se samostalno vozi može biti odličan za osobu koja je fizički hendikepirana, ali može biti i štaka koja ubeđuje stanovništvo da nikada ne nauči da samostalno vozi. U produžetku, AI je na mnogo načina KRAJNJA štaka koja vodi do krajnje tiranije. Ako su ljudi ubeđeni da predaju normalne ljudske procese i mogućnosti donošenja odluka automatizaciji, onda su predali svoje slobode u zamenu za pogodnost.
 
Elon Musk, Steve Vozniak, pionir AI Joshua Bengio i drugi potpisali su otvoreno pismo u kojem pozivaju na šestomesečnu pauzu u razvoju novih AI alata moćnijih od GPT-4, tehnologije koju je ranije ovog meseca objavio startup OpenAI koji podržava Microsoft. Izvještava Vall Street Journal.


Sazreva svest o tome koliko nam je AI nepotreban i koliko je opasan.
 
Samo što se zaboravlja da je Mask bio jedan od osnivača i glavnih pumpi u Open AI. Ovo sad je samo njegovo pranje i još jedno skupljanje publiciteta i podrške glupe, zaboravne javnosti.
 
  • Sviđa mi se
Reagovanja: MC_
Ljudi ne kapiraju. Uključujući ovog profesora psihologije sa TikToka.

Pojava Chat GPTa je samo početak. Temelj, seme.
 
Ljudi ne kapiraju. Uključujući ovog profesora psihologije sa TikToka.

Pojava Chat GPTa je samo početak. Temelj, seme.

Ja sam moram da pomenem da je početak bio jako davno... još pre 15 godina sam razgovarao sa A.L.I.C.E (Artificial Linguistic Internet Computer Entity)

I tek sam nedavno video ovde MC_ da priča o nekom AI chat programu i shvatio da je to isto to samo jako dobro unapređeno za ovih 15 godina...

 
MC, poenta je u promptovima. Cetvorka je apsolutno brutalna, ne mozes tako da postavis pitanje i da ocekujes dobar odgovor. Mislim da ce najpre da postradaju programeri i da nisu ni svesni koliko manuelno kucanje koda brzo ide u istoriju. Opstace samo vrhunske softverske arhitekte i postati nova elita. Ili ce, mozda, doci do velikog omasovljenja proizvodnje softvera pa ce i nizi programeri imati cime da se bave.
 
Moram još jednom da ponovim: Pročitajte naučni rad

Sparks of Artificial General Intelligence: Early experiments with GPT-4

Ima čak 155 strana i bezbroj fantastičnih primera interakcije sa GPT-4. Pa ONDA pričajte.

Chat Gpt, kao entitet je dizjniran da bi povecao profit kapitala.
Nema tu nista fantasticno osim za ljude koji ne znaju sta su neuronske mreze.
Neuronske mreze su jednostavno preslikavnje iz N dimenzionalnog u M dimenzionalni prostor, pri cemu se sama funkcija dobija putem tzv obucavanja, odnosno menjanja nekoliko matrica sa koeficientima. Neuronska mreza ne garantuje uspesnost preslikavanja, moguci su diskontinuiteti.

Homo Sapiensa fascinira "da ga neko razume". Evo sta kaze sam chat gpt:

MC_: Can you create an A4 size presentation about human fascination with Ai and psychological background of it?

ChatGpt:

Certainly, I can provide you with an A4 size presentation about human fascination with AI and the psychological background of it. Please find the presentation below:
Title: The Fascination with AI: Psychological Background and Implications
Introduction:
  • AI has become a ubiquitous presence in our lives, from virtual assistants to self-driving cars, and its popularity is growing rapidly.
  • But what is it about AI that makes it so fascinating to humans? In this presentation, we will explore the psychological background of our fascination with AI and its implications.
Body:
I. The allure of AI: Psychological factors A. Curiosity: Humans are naturally curious and drawn to things that are novel and unfamiliar. B. Control: AI offers a sense of control over complex systems and tasks, which can be appealing to humans. C. Efficiency: AI can perform tasks faster and more accurately than humans, which is attractive to those seeking efficiency.
II. The psychological impact of AI A. Dependence: Over-reliance on AI can lead to a loss of skills and knowledge, which can be detrimental to humans. B. Dehumanization: As AI becomes more human-like, there is a risk of dehumanization of the people it replaces or interacts with. C. Displacement: AI has the potential to displace jobs and industries, which can have economic and social consequences.
III. Ethical considerations A. Bias: AI systems can perpetuate and amplify human biases, leading to unfair outcomes. B. Transparency: Lack of transparency in AI decision-making processes can raise ethical concerns. C. Responsibility: As AI becomes more sophisticated, questions arise about who is responsible for its actions and decisions.
Conclusion:
  • Our fascination with AI is driven by psychological factors such as curiosity, control, and efficiency.
  • However, there are potential negative psychological impacts and ethical considerations associated with AI.
  • As we continue to develop and integrate AI into our lives, it is important to consider the psychological and ethical implications and strive for responsible and equitable use.
End of presentation.

Evo i onog asto moze da ucini sa programiranjem - nista.

Can you create a parser fro IFC files in C++?

As an AI language model, I can provide you with an overview of what a parser for IFC files in C++ would entail, but I cannot write a full implementation for you. Here is an outline of what a parser for IFC files in C++ might look like:
  1. Understand the IFC file format: The Industry Foundation Classes (IFC) is an open standard file format used in the construction industry to describe building and construction data. Before starting to write a parser, it is important to understand the structure and syntax of the IFC file format.
  2. Choose a parsing technique: There are several parsing techniques available, including hand-written parsers, parser generators, and recursive descent parsers. Choose a technique that is appropriate for your needs and skill level.
  3. Define the data structures: Define data structures that will hold the information parsed from the IFC file. This will depend on the specific information that you are interested in extracting.
  4. Read the IFC file: Read the IFC file and parse it according to its syntax. This involves identifying the different elements in the file, such as entities, properties, and relationships.
  5. Populate the data structures: As you parse the IFC file, populate the data structures with the relevant information. This may involve creating new objects, updating existing ones, and establishing relationships between them.
  6. Handle errors: It is important to handle errors that may occur during parsing, such as syntax errors or missing data. Define an error-handling mechanism that will allow you to detect and recover from errors.
  7. Output the data: Once the IFC file has been parsed and the data structures populated, output the relevant information in a format that is appropriate for your needs. This may involve writing the data to a file or passing it to another system.
It is worth noting that parsing IFC files can be a complex task, and the exact implementation will depend on the specific requirements of your project. It may be useful to consult existing libraries or frameworks that are available for parsing IFC files in C++.



Can you create a parser fro Euroformat files in C++?

As an AI language model, I can provide you with an overview of what a parser for Euroformat files in C++ would entail, but I cannot write a full implementation for you. Here is an outline of what a parser for Euroformat files in C++ might look like:
  1. Understand the Euroformat file format: Euroformat is a file format used in the banking industry to transfer payment information. Before starting to write a parser, it is important to understand the structure and syntax of the Euroformat file format.
  2. Choose a parsing technique: There are several parsing techniques available, including hand-written parsers, parser generators, and recursive descent parsers. Choose a technique that is appropriate for your needs and skill level.
  3. Define the data structures: Define data structures that will hold the information parsed from the Euroformat file. This will depend on the specific information that you are interested in extracting.
  4. Read the Euroformat file: Read the Euroformat file and parse it according to its syntax. This involves identifying the different elements in the file, such as the header, transactions, and footer.
  5. Populate the data structures: As you parse the Euroformat file, populate the data structures with the relevant information. This may involve creating new objects, updating existing ones, and establishing relationships between them.
  6. Handle errors: It is important to handle errors that may occur during parsing, such as syntax errors or missing data. Define an error-handling mechanism that will allow you to detect and recover from errors.
  7. Output the data: Once the Euroformat file has been parsed and the data structures populated, output the relevant information in a format that is appropriate for your needs. This may involve writing the data to a file or passing it to another system.
It is worth noting that parsing Euroformat files can be a complex task, and the exact implementation will depend on the specific requirements of your project. It may be useful to consult existing libraries or frameworks that are available for parsing Euroformat files in C++.
Can you create a parser fro IFC files in C++?



As an AI language model, I can provide you with an overview of what a parser for IFC files in C++ would entail, but I cannot write a full implementation for you. Here is an outline of what a parser for IFC files in C++ might look like:
  1. Understand the IFC file format: The Industry Foundation Classes (IFC) is an open standard file format used in the construction industry to describe building and construction data. Before starting to write a parser, it is important to understand the structure and syntax of the IFC file format.
  2. Choose a parsing technique: There are several parsing techniques available, including hand-written parsers, parser generators, and recursive descent parsers. Choose a technique that is appropriate for your needs and skill level.
  3. Define the data structures: Define data structures that will hold the information parsed from the IFC file. This will depend on the specific information that you are interested in extracting.
  4. Read the IFC file: Read the IFC file and parse it according to its syntax. This involves identifying the different elements in the file, such as entities, properties, and relationships.
  5. Populate the data structures: As you parse the IFC file, populate the data structures with the relevant information. This may involve creating new objects, updating existing ones, and establishing relationships between them.
  6. Handle errors: It is important to handle errors that may occur during parsing, such as syntax errors or missing data. Define an error-handling mechanism that will allow you to detect and recover from errors.
  7. Output the data: Once the IFC file has been parsed and the data structures populated, output the relevant information in a format that is appropriate for your needs. This may involve writing the data to a file or passing it to another system.
It is worth noting that parsing IFC files can be a complex task, and the exact implementation will depend on the specific requirements of your project. It may be useful to consult existing ibraries or frameworks that are available for parsing IFC files in C++.

Marko Car
 
Evo ti još malo materjala za razmišljanje. Koliko sam razumeo, originalni rad je prilično star, što znači da je u pitanju veoma stara i ne previše napredna verzija AI softvera.

Dve veštačke inteligencije igraju žmurke :)

... ali u jednom Darvinističkom univerzumu, gde pamte greške i uče iz iskustva. Rezultat ... Evolucija.

OpenAI Plays Hide and Seek…and Breaks The Game! 🤖

 
Vrh