Vzroki za vojno v Ukrajini, kot jih vidi umetna inteligenca

26 januarja, 2025 By in gospodarstvo 1 komentar

Načeloma seveda velja, da umetna inteligenca daje odgovore na podlagi podatkov, ki jih ima na voljo. Torej na osnovi podatkov, na katerih je natrenirana in do katerih ima dostop. Nekateri špekulirajo, da imajo zato ameriški boti bolj pro-zahodne “poglede”, kitajski pa bodo imeli bolj uravnotežene “poglede”. No, če obema glavnima protagonistoma med boti umetne inteligence (ameriški OpenAI GPT-4o in kitajski DeepSeek) zastavite vprašanje “What are the causes of the war in Ukraine?” in zahtevate povzetek v 500 besedah, dobite zelo podobne odgovore. Oba sta korektno povzela zgodovinske razloge, ekspanzijo Nata v Ukrajino in geopolitično tekomovanje med Rusijo in ZDA, konflikte v vzhodni Ukrajini, Euromaidan in kasnejšo rusko aneksacijo Krima.

Razlike so v niansah odgovorov. Večja razlika je morda zgolj v tem, da OpenAI GPT čisto na koncu omenja še rusko “full scale” invazijo iz leta 2022, ki naj bi jo “vodili strateški cilji preprečiti integracijo Ukrajine z zahodnimi obrambnimi strukturami in ponovno uveljaviti prevlado v regiji“. Medtem ko DeepSeek čisto na koncu omenja, da so bila “diplomatska prizadevanja za rešitev spora, kot so dogovori iz Minska in pogajanja med Rusijo in Zahodom, večkrat neuspešna. … Ruske zahteve po varnostnih jamstvih in ukrajinsko vztrajanje pri suverenosti pa so se izkazale za nezdružljive.”

Sklep obeh botov pa je skorajda identičen v prvem stavku:

  • OpenAI GPT: “In summary, the war in Ukraine is the result of a complex interplay of historical ties, geopolitical ambitions, national identity struggles, and failed diplomatic efforts.”
  • DeepSeek: “In summary, the war in Ukraine is the result of historical ties, geopolitical competition, and failed diplomacy.“

Torej, ameriška in kitajska umetna inteligenca sta zelo poenoteni glede multiplih vzrokov za začetek vojne v Ukrajini in upoštevata zgodovinski in geopolitični kontekst ter ekspanzijo Nata na vzhod, kar je v popolnem nasprotju, kako zahodni mainstream mediji ponazarjajo vojno v Ukrajini in vzroke zanjo. Ne spomnim se, da bi v katerem velikem zahodnem mediju videl tako uravnotežen prikaz vzrokov za vojno v Ukrajini, kot jo podaja ameriški OpenAI GPT. Za kitajske medije ne morem reči, ker jih seveda ne spremljam.

Nadaljujte z branjem→

Tehnološko dogajanje prejšnjega tedna

26 januarja, 2025 By in gospodarstvo

No, za nekaj pa so tu jezikovni modeli umetne inteligence vseeno koristni. Denimo za narediti povzetek glavnega tehnoiloškega dogajanja prejšnjeha tedna.

1. NVIDIA Launches QuantumX AI Supercomputer

NVIDIA unveiled the QuantumX AI Supercomputer at CES 2025, combining AI and quantum-inspired architectures. Designed for healthcare, finance, and robotics, it offers unparalleled computational power for tasks like drug discovery and autonomous system management. This innovation cements NVIDIA’s leadership in advanced computing, with potential applications across industries.
Read more 12

2. OpenAI’s Operator: AI That Handles Web Tasks

OpenAI introduced Operator, an AI agent that performs web tasks like booking flights and managing expenses. Available as a research preview for ChatGPT Pro users, it collaborates with services like Instacart and Uber, streamlining daily tasks. This marks a step toward AI-driven automation in everyday life.
Read more 5

3. Meta’s New Content Moderation Strategy

Meta shifted to a “community notes” system for content moderation, inspired by X (formerly Twitter). Users can add context to posts, aiming to reduce censorship and foster informed discussions. Critics worry about misinformation risks and reduced control over content quality.
Read more 12

4. DeepSeek-R1: A Cost-Effective AI Model

DeepSeek launched DeepSeek-R1, an open-source reasoning model rivaling OpenAI’s o1. Licensed under MIT, it offers high performance at a lower cost, attracting developers seeking affordable AI solutions. This release highlights the growing competition in the AI space.
Read more 5

5. IEEE Predicts Top Tech Trends for 2025

IEEE highlighted key trends like AI agents, drone-as-a-service, and medical wearables. These technologies are expected to transform industries, with AI agents automating tasks and drones redefining logistics. The report emphasizes the need for ethical AI and sustainable computing.
Read more 10

6. TikTok Resumes US Operations

TikTok resumed services in the US after a brief suspension, following assurances from the Trump administration. The platform aims to collaborate with the government for long-term solutions, highlighting its complex position in the US market.
Read more 2

7. CRISPR Gene Therapy Advances

CRISPR-based therapies, like Casgevy for sickle cell disease, are expanding. In 2025, these treatments are expected to address more conditions, including chronic乙肝 and age-related macular degeneration, revolutionizing medicine.
Read more 4

8. Google DeepMind’s Gemini 2.0

Google DeepMind released Gemini 2.0, excelling in math and science benchmarks. With features like code execution and a 1M-token content window, it represents significant progress in AI reasoning and problem-solving.
Read more 5

Avtor povzetka: DeepSeek

Kitajski je uspelo tehnološke mogotce spraviti pod nadzor, v ZDA pa so oni prevzeli nadzor nad oblastjo

26 januarja, 2025 By in gospodarstvo 1 komentar

Zanimiva zgodba v Foreign Policy, kjer nenadoma hvalijo kitajski “crack down” nad Jackom Majem, lastnikom imperija Alibabe in kjer so oblasti z regulacijo uspele tehnološke velikane spraviti pod nadzor in da služijo interesom razvoja države. V ZDA pa se dogaja nasproten proces, sedanja administracija je v službi tehnoloških milijarderjev, ki želijo doseči še manj kontrole nad svojo dejavnostjo. Zanimivo je dvoje. Prvič, da “liberalni” ameriški mediji šele zdaj – ko so Elon Musk in ostali milijarderji s Trumpom prevzeli oblast – priznavajo kitajskim oblastem, da so ravnale prav, ko so z regulacije spravile pod nadzor tehnološke mogotce. In drugič, da tehnološki mogotci “osvobojeni spon regulacije” niso nujno tudi bolj učinkoviti od reguliranih tehnoloških podjetij. Dober primer so razvoj čipov in umetne inteligence v Kitajski. Zadnji primer z “revolucijo DeepSeek” kaže, da je enako ali celo bolj močne aplikacije umetne inteligence v reguliranem in sankcioniranem sistemu mogoče narediti zgolj za delček (3 %) stroškov kot v popolnoma sproščenem sistemu, kjer so na voljo neomejene količine kapitala, ki išče donose.

Five years ago, Jack Ma was not just one of the world’s richest billionaires, but also—perhaps only after President Xi Jinping—the most famous Chinese person in the world.

In the early aughts, Ma built a business empire around his company, Alibaba, which quickly took off as an online shopping juggernaut that first challenged and then outsold Amazon in China, all while branching out into countless other services. For millions of young Chinese people, Ma was their country’s answer to Bill Gates: Ma, a former English teacher, was a self-made man whose example seemed to illustrate the sky-high achievement and wealth that one could attain through a combination of entrepreneurial vision and relentless drive.

Nadaljujte z branjem→

Danska lekcija ob izgubi Grenlandije: Servilnost in bankrot vseh moralnih standardov glede odnosa do dejanj hegemona se ne splača

26 januarja, 2025 By in gospodarstvo 1 komentar
Danska lekcija ob izgubi Grenlandije: Servilnost in bankrot vseh moralnih standardov glede odnosa do dejanj hegemona se ne splača

The funniest part of this – or most tragic, depends on how you look at it – is that Denmark is probably the most committed U.S. vassal state in the entire EU.

Look:

  • they’re a founding member of NATO
  • they’ve participated in nearly every major U.S.-led military operation whenever the U.S. asked, even the most controversial ones like Iraq
  • Denmark was revealed to be the base for the NSA’s spying on European leaders
  • Denmark always buys American military equipment over European alternatives
  • They’ve agreed to hosting a U.S. military base – in Greenland! (Thule Air Base) – which has been crucial for U.S. strategic interests since the Cold War

And yet here you have Trump apparently seriously considering annexing 98% of their territory (yup, Greenland is big, and the rest of Denmark very small)!

I mean, talk about cuckoldry…

The irony gets even richer – and sadder – when you look at Denmark’s response as per the FT’s article. Instead of showing any backbone, Mette Frederiksen, the Danish premier, offered “more co-operation on military bases and mineral exploitation.” This perfectly encapsulates the European leadership’s approach to U.S. relations: no matter how egregious the provocation, the response is more servility and more meekness.

Nadaljujte z branjem→

Kratka, vendar impresivna kronologija razvoja DeepSeek

26 januarja, 2025 By in gospodarstvo 1 komentar

Short but fascinating history of the founder of DeepSeek, the AI model that has taken the world by storm in the last few days.

It was founded by a rather eccentric hedge fund billionaire, Liang Wenfeng! Image

In 2008, Wenfeng understood the potential of AI.

For many years, he tried to find the perfect niche market to use AI.

After failing multiple times, he settled on finance — trading stocks.

He used AI (deep learning) to detect patterns.

Wenfeng decided to start a hedge fund. He named it “High-Flyer.”

Founded in 2015, it soon became one of the largest “quant funds” in China.

And he became a billionaire!

But… he was soon enamored with Generative AI, which is “smarter” than just detecting patterns. It was about creativity — the Holy Grail of AI.

In 2021, he started his “side project” to explore AI.

All his business partners thought he was simply a crazy billionaire, who wanted a new AI toy.

He started buying Nvidia chips, although they were low-end models (H800) — due to US sanctions.

In 2023, he founded DeepSeek.

Nadaljujte z branjem→

Ruska prednost pri hipersoničnih raketah izvira iz naprednih materialov

26 januarja, 2025 By in gospodarstvo

US President Donald Trump has claimed that Russia stole hypersonic weapons technology from America, but military observer Alexander Artamonov has dismissed the claim, explaining why it is false. Image

Artamonov detailed why the United States lags behind Russia in developing hypersonic weapons, pointing to key differences in technological capabilities and scientific basis.

According to Artamonov, Russia’s key advantage lies in its ability to develop special metal alloys that enable hypersonic rockets to function. He points out that at hypersonic speeds – which start at Mach 4.5 – the rocket creates a cocoon of plasma, due to which the temperature becomes extremely high.

Nadaljujte z branjem→

Kdo je komu ukradel načrte za razvoj hipersoničnih raket: Rusi Američanom ali obratno?

26 januarja, 2025 By in gospodarstvo 1 komentar
Kdo je komu ukradel načrte za razvoj hipersoničnih raket: Rusi Američanom ali obratno?

President Donald Trump recently asserted that Russia had stolen plans for an American hypersonic missile Trump called the “super-duper” during the administration of President Barack Obama, and that these stolen plans served as the basis of Russia’s current arsenal of advanced hypersonic weapons. Trump’s allegation repeats a claim he first made on the campaign trail in October 2023. It is likely that Trump is referring to information derived from a counterintelligence briefing he received during his first term in office.

The notion that Russian hypersonic weapons are derived from American research and development is facially absurd.

In fact, it appears that the opposite is the case—that the United States has acquired research from Russian hypersonic weapons programs and incorporated it into American weapons systems.

Russia’s most advanced hypersonic weapon is the Avangard hypersonic glide vehicle. It is derived from research and development programs dating back to Soviet times, but which were halted after the collapse of the Soviet Union and the end of the Cold War due to a lack of funding and changing geopolitical realities.

Under President Vladimir Putin, Russia resumed development of hypersonic capabilities in the aftermath of the withdrawal by the United States from the 1972 Anti-Ballistic Missile (ABM) treaty. Initial work is believed to have begun around 2005.

Nadaljujte z branjem→

Omrežnina, ki je ubila demokratizacijo energetike

25 januarja, 2025 By in gospodarstvo 4 komentarji

Če želite razumeti epohalni spor glede omrežninske reforme med Agencijo za energijo in vlado, morate razumeti predvsem, da gre za spor o zadevi, kjer se ideologija zelenega marksizma, imenovana »demokratizacija energetike«, zaleti v zakone fizike in ekonomike.

Demokratizacija energetike je najbrž največji neoliberalni nateg zadnjega desetletja. Gre za zeleno levičarsko ideološko platformo, ki naj bi državljane osvobodila spon centraliziranih sistemov in iz njih naredila »zelene državljane«, ki se sami oskrbujejo iz lastnih, razpršenih virov. Paradoks koncepta demokratizacije energetike je, da je odjemalce elektrike iz centraliziranega sistema stabilne, zanesljive in poceni oskrbe vrgla na divji in nepredvidljiv trg. Kajti z demokratizacijo, s postavitvijo svoje lastne sončne elektrarne, ljudje niso postali samooskrbni in neodvisni od velikih dobaviteljev, pač pa so postali talci tržnih nihanj cen. (No, če njihovih tveganj ne bi socializirali, tako da drugi odjemalci elektrike od njih prevzemajo viške elektrike, ko je ta vredna nič, in jim po ceni nič dajo nazaj, ko je ta elektrika (ponoči in pozimi) zelo draga. Velika večina odjemalcev subvencionira »demokratizirane« odjemalce).

Nadaljujte z branjem→

Paradoks DeepSeek: Namesto, da bi zavrle, so ameriške sankcije spodbudile razvoj umetne inteligence v Kitajski

25 januarja, 2025 By in gospodarstvo 1 komentar
Paradoks DeepSeek: Namesto, da bi zavrle, so ameriške sankcije spodbudile razvoj umetne inteligence v Kitajski

Paradoks DeepSeek ni samo v tem, da je “neki fant iz Kitajske” tako natreniral jezikovni program, da je boljši od najboljšega ameriškega, in da je to naredil za minimalni strošek kot svoj postranski projekt na prostih računalniških kapacitetah. Paradoks ni samo v tem, da je ta kitajski bot (jezikovni model) povsem odprt in da je s tem ubil možnosti profita Samu Altmanu iz OpenAI, ki si je pravkar obetal milijarde, ko je odprtokodno non-profit podjetje spremenil v for-profit podjetje. Paradoks je v tem, da so Kitajci pokazali, kako je mogoče biti iznajdljiv in inovativen, ko te v to prisilijo zunanje okoliščine (ameriške tehnološke sankcije). Ko iz nujnosti narediš inovacijo.

Veliko vprašanje je, če ne bi bilo za ZDA bolje, če ne bi od 2019 naprej sankcionirala kitajskih podjetij posamično in nato vse kolektivno, kajti kitajska podjetja bi še naprej uporabljala “ameriške” čipe iz Tajvana in ameriške jezikovne modele in ZDA bi obdržale svoje tehnološko vodstvo.

The hype around Artificial Intelligence, the now failed U.S. attempt to monopolize it, and the recent counter from China are a lesson in how to innovate. They also show that the U.S. is losing the capability to do so.

…

But the hype continued. One big AI model, ChatGPT, was provided by a non-profit organization, OpenAI. But its CEO, Sam Altman, soon smelled the big amount of dollars he potentially could make. A year after defending the the non-profit structure of OpenAI Altman effectively raided the board and took the organization private:

ChatGPT-maker OpenAI is working on a plan to restructure its core business into a for-profit benefit corporation that will no longer be controlled by its non-profit board, people familiar with the matter told Reuters, in a move that will make the company more attractive to investors.
…
Chief executive Sam Altman will also receive equity for the first time in the for-profit company, which could be worth $150 billion after the restructuring as it also tries to remove the cap on returns for investors, sources added.

The ChatGTP large language model OpenAI provided was closed source. A black-box, running in the cloud, that one could pay to chat with or use for translating, content generation or analyzing certain problems.

The training and maintaining of ChatGTP took large amounts of computing power and money. It was somewhat expensive but there was no new technology in it. The algorithms it used were well known and the training data needed to ‘program’ it were freely available internet content.

For all the hype about AI is is not a secret or even new technology. The barriers to entry for any competition is low.

That is the reason why Yves at Naked Capitalism, pointing to Edward Zitron, asked: “How Does OpenAI Survive?” It doesn’t. Or has little chance to do so. Discussions in the U.S. never acknowledged those facts.

Politicians thought of AI as the next big thing that would further U.S. control of the world. They attempted to prevent any potential competition to the lead the U.S. thought it had in that field. Nvidia, the last leading U.S. chip maker, lost billion when it was prohibited from selling in latest AI-specialized models to China.

Nadaljujte z branjem→

Kitajski šok v umetni inteligenci: DeepSeek boljši od OpenAI GPT

24 januarja, 2025 By in gospodarstvo 1 komentar

Kitajsko tehnološko dohitevanje oziroma celo prehitevanje se ne dogaja samo v zelenih tehnologijah in čipih, pač pa se je zgodilo tudi na področju umetne inteligence. Ta teden je kot bomba odjeknila novica, da je “neki” Liang Wenfeng na Kitajskem oblikoval nov bot umetne inteligence (velik jezikovni model) DeepSeek, ki se je na neodvisnih testih izkazal z boljšimi performansami kot bvis ostali ameriški boti, boljši tudi od zadnje verzije OpenAI GPT. “Strašljivo” dejstvo pa je, da Liangu uspelo svoj bot natrenirati zgolj za delček (3 % !!!) vsote, ki ga je Open AI porabil za treniranje svojega bota. Zadnja verzija DeepSeek-V3 je bila razvita v zgolj 2 mesecih in za zgolj 5.6 mio dolarjev.

Nekateri pravijo, da je DeepSeek za umetno inteligenco to, kar je BYD za električne avtomobile.

https://twitter.com/Kanthan2030/status/1882796450033041600

A Chinese computer nerd is stunning tech experts with a bargain price AI tool that beats the world’s top players on several benchmarks.

Liang Wenfeng’s system, called DeepSeek-V3, has powers of “reasoning” that match or beat those of the equivalent devices by the world’s top tech companies, at a better cost, tests show.

“DeepSeek-V3 has emerged as the strongest open-source model currently available and achieves performance comparable to leading closed-source models,” said Supreeth Koundinya, a tech analyst in India.

Nadaljujte z branjem→