Umetna inteligenca

Hej, hvala za povratno. Razumem poanto: vsak lahko napiše nekaj v LLM in dobi odgovor.
Moja poanta je druga: agentni sistem ? "tipkam namesto tebe v chatbot".

Kar delam jaz, ni posredništvo, ampak orkestracija:

jasno definiran cilj -> orodja/akcije (iskanje, računanje, klici API),

pravila in varovalke (shema, validacija, "don't-know" prag, HITL),

spomin/kontekst in ponovljiv workflow,

ter merjenje (latenca, natančnost, strošek), da je zadeva uporabna za ekipo, ne le za en odgovor.

Če želiš, predlagaj 3 konkretne naloge (npr. "preberi 5 virov, izračunaj X, pripravi output v JSON po shemi in pošlji v orodje Y").
Dostavim repro (skripta + logi + rezultat), da vidimo razliko med "vtipkam v LLM" in agentnim procesom, ki je zanesljiv in merljiv.

Ne želim smetiti foruma - z veseljem pokažem na primeru. Če se motim, bom to tudi priznal. Deal? ?

Ma men se zdi kar uporabna. Sicer sem vedel že rešitev ampak sem šel vprašat.(Spraševal sem okoli Spark ade in Rušta kjer se dbc(design by contract) realizira prek makrov) Pa izvem med tem da obstajajo tudi negativne temperature na Kelvinovi lestvici.

Jah nekaj 1000x bolj uporabno kot išlo-tech.

Če tukaj vprašaš kaj tahničnega si najprej dva med sabo jebeta mater potem smo pa na koncu pri Rusih ukraincih trumpu golobu ...

@David Mayer hehe, tale AI res odpira nova vesolja - od negativnih kelvinov do tega, da brez njega ne znamo več živeti :D Kyrca s 5-urno sejo je sicer fajn, dokler se ne raztegne kot žvečilni. Sem pa za solidarnost pri uporabi, da ne bo treba šaltat na enterprise, pa bo še za n00be dovolj :)

No ja. Jaz sem s svojo freechad verzijo prišel do izračuna dejanskega odvodnjavanja meteorne vode s strehe. Res je trajalo ampak, če veš kaj želiš ti UI kar pomaga. Če pa ne veš in ji zaupaš prvi odgovor, pa je lahko zate katastrofa.
Če se omejim na tehnične probleme, bi LLM moral najprej s svojimi vprašanji ugotoviti, kdo jo za kaj sprašuje. Ampak to seveda pomeni, da se izjasniš, kakšno izobrazbo imaš ali pa seveda vprašanje postaviš svoji izobrazbi primerno.

@David Mayer mislim, da smo se vsi sprehodili skozi vse ChatGPT nivoje - od free do pro pa do enterprise s "svobodo gibanja" ? Razlike so, ampak če znaš pravilno zastaviti vprašanje in pripraviš dober prompt, lahko iz LLM-ja izvlečeš več kot samo 3D model sončnega sistema. Tudi "Hello World" projekti z lunicami se razvijejo v kaj večje in bolj koristnega, če jim daš malo hrane in potrpljenja.

@mirator tvoja anekdota o izračunu meteorne vode s strehe je super! AI res pomaga, če veš, kaj iščeš, in na kakšen način ti mora razložiti - zato ni slabo, če se včasih pohvalimo s svojo izobrazbo in ciljem ? Pomembno je, da ostanemo radovedni, povprašamo večkrat in preverimo odgovore, sicer nam namesto odtokov naredi vodni tobogan!

@David Mayer Haha, enterprise CMS/ERP/CRM v treh urah - to je skoraj tako hitro kot Hello World ? Specializirani modeli imajo svoje mesto, sploh ko gre za optično prepoznavo LEGO kock; generalist kot GPT pa je super za navdih, pisanje specifikacij ali iskanje analogij med modulom in kocko. In res je, pri Lego analizi ne smemo pozabiti na prave čepke in osvetlitev - drugače bo model videl samo nered. Ni pa slabo imeti oba svetova: enega za natančne čačke in drugega za kreativni kaos!

Jaz chatGTP uporabljam predvsem takrat, ko vem kakšen rezultat bi moral dobiti, ne vem pa kako do njega priti. In kot sem omenil za meteorno vodo, sem s pomočja ChatGTP prišel tudi do drugih verodostojnih rešitev.Seveda ChatGTP rabi tudi tvojo pomoč predvsem v smislu razumevanja osnovnega vprašanja.
Torej, če veš, kaj pričakuješ, je Chat(katerikoli) izredno uporabno orodje nasprotno, če ne veš kakšen mora biti rezultat, je lahko odgovor Chata(xy) izreno nevaren.

Gregor P je 9. okt 2025 ob 15:30:01 izjavil:

@David Mayer strinjam se – razlike med free, pro in enterprise se poznajo, a z dobrim, natančnim promptom lahko iztisneš ogromno tudi iz osnovnejših nivojev. Dolgi, podrobni opisi zahtev so v bistvu že specifikacije, iz katerih LLM lahko generira vse, od projektne dokumentacije do 3D vizualizacije sončnega sistema s planeti, lunami, konfiguracijskimi paneli in svobodo gibanja 😄

Malo časa je vzelo od A do B, kje je teleportacija, ko jo človek res potrebuje? Včasih preprosto 'kapacitete' zmanjka, tudi tedenske, saj zato pa tiščijo v enterprise, ki je 5x dražji. Saj je še več zanimivosti, samo naj ostane še kaj za drugič, stvari na pladnju žal niso vedno cenjene.

Ja, je animacija. Sam je za pogledat

Evo ... s tem Frčem sem že imel eno zanimivo debato:

While Intelligence is becoming Artificial, Stupidity will always remain Original.
(vir X)

Prosim - verjetno sem eden bolj redkih, ki sam ne gledam videov, ker so mnogi zamudni in švoh v informacijskem smislu s preveč balasta in vtkanih reklam... Ostali pač glejte, kakor želite. Včasih sicer pogledam diy/work in diy/work fails, road rage, občasno še kaj o navtiki, potapljanju, aviatiki, potem pa se počasi neha, zelo dolgočasno, predvidljivo in očitno precej nezanimivo, primerjujoč števila ogledov.

Here are the summaries in 10, 3, and 1 sentence:

**10 sentences:**
The video criticizes OpenAI and its CEO, Sam Altman, for releasing Sora 2, a TikTok knockoff app that generates AI slop videos. The app is seen as a desperate attempt to keep the AI hype cycle going and attract investors. The video highlights several issues with Sora 2, including its potential for generating deepfakes, copyright infringement, and racist content. The app's economics are also questioned, as it costs OpenAI massive amounts of money to generate videos. The video argues that the AI industry is in a bubble, with investors pouring billions of dollars into companies that are not generating significant revenue. The bubble is fueled by hype and the promise of "super intelligence" or AI that is smarter than humans. However, the video questions whether this is even possible and notes that current AI models are not getting significantly better. The video concludes that the AI industry is likely to collapse, causing significant economic harm to regular people. The speaker argues that the tech investors know that they are building a bubble and will profit from it, while regular people will suffer. The video ends with a call to action to use human brains and stop listening to the hype surrounding AI.

**3 sentences:**
The video criticizes OpenAI and its CEO, Sam Altman, for releasing Sora 2, a TikTok knockoff app that generates AI slop videos, which is seen as a desperate attempt to keep the AI hype cycle going. The video highlights several issues with Sora 2, including its potential for generating deepfakes and copyright infringement, and questions the app's economics. The speaker argues that the AI industry is in a bubble, fueled by hype and the promise of "super intelligence," and is likely to collapse, causing significant economic harm to regular people.

**1 sentence:**
The video criticizes OpenAI and its CEO, Sam Altman, for releasing Sora 2, a TikTok knockoff app that generates AI slop videos, and argues that the AI industry is in a bubble that is likely to collapse, causing significant economic harm to regular people.

Here are the summaries in 10, 3, and 1 sentence:

**10 sentences:**
The video discusses the state of the AI industry and whether it is a bubble. The speaker presents various opinions from experts and analysts, ranging from those who believe it is not a bubble to those who think it is a huge bubble. Jim Cramer and Mark Zuckerberg are cited as believers in the AI industry, while Hank Green and Ed Zitron are among those who question the amount of investment. Paul Tudor Jones suggests that if it is a bubble, it is a small one, while another analyst claims it is 17 times the size of the dot-com bubble. The speaker notes that most people agree that the industry is frothy, but disagree on the severity of the situation. The video concludes that nobody really knows what the future holds, and timing the market is difficult. The speaker summarizes the various perspectives on the AI bubble and notes that most people think it is a bubble of some kind. The video ends with the speaker acknowledging that the industry is likely overinvested, but the extent of the problem is unclear. The speaker's goal is to provide a summary of the different perspectives on the AI bubble.

**3 sentences:**
The video discusses the state of the AI industry and whether it is a bubble, presenting various opinions from experts and analysts. While some believe it is not a bubble, others think it is a huge bubble, with estimates ranging from a small bubble to 17 times the size of the dot-com bubble. The speaker concludes that most people agree that the industry is frothy, but disagree on the severity of the situation.

**1 sentence:**
The video summarizes various perspectives on whether the AI industry is a bubble, with opinions ranging from it not being a bubble to being a huge bubble, and concludes that while most people agree it's frothy, the extent of the problem is unclear.

Saj menda lahko uporabljamo isto orodje?

openai je nov ms/google ?
Delajo spremembo njihovega statusa podjetja (iz non profit), nvida je vrgla 100 milijard notri, kmalu bodo začeli ponujati plačljive produkte za vse živo.