Skip Navigation
InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)HO
Hotznplotzn @lemmy.sdf.org
Posts 27
Comments 15
www.bbc.com Australia bans DeepSeek on government devices over security risk

The Australian government says the Chinese AI app is a threat to it and its assets.

Australia bans DeepSeek on government devices over security risk

cross-posted from: https://lemmy.sdf.org/post/28980041

> Australia has banned DeepSeek from all government devices and systems over what it says is the security risk the Chinese artificial intelligence (AI) startup poses. > > ... > > Growing - and familiar - concerns > > Western countries have a track record of being suspicious of Chinese tech - notably telecoms firm Huawei and the social media platform, TikTok - both of which have been restricted on national security grounds. > > ... > > An Australian science minister previously said in January that countries needed to be "very careful" about DeepSeek, citing "data and privacy" concerns. > > The chatbot was removed from app stores after its privacy policy was questioned in Italy. The Italian goverment previously temporarily blocked ChatGPT over privacy concerns in March 2023. > > Regulators in South Korea, Ireland and France have all begun investigations into how DeepSeek handles user data, which it stores in servers in China. > > ... > > Generally, AI tools will analyse the prompts sent to them to improve their product. > > This is true of apps such as ChatGPT and Google Gemini as much as it is DeepSeek. > > All of them gather and keep information, including email addresses and dates of birth. > > ...

0
Technology @midwest.social Hotznplotzn @lemmy.sdf.org
www.bbc.com Australia bans DeepSeek on government devices over security risk

The Australian government says the Chinese AI app is a threat to it and its assets.

Australia bans DeepSeek on government devices over security risk

cross-posted from: https://lemmy.sdf.org/post/28980151

> cross-posted from: https://lemmy.sdf.org/post/28980041 > > > Australia has banned DeepSeek from all government devices and systems over what it says is the security risk the Chinese artificial intelligence (AI) startup poses. > > > > ... > > > > Growing - and familiar - concerns > > > > Western countries have a track record of being suspicious of Chinese tech - notably telecoms firm Huawei and the social media platform, TikTok - both of which have been restricted on national security grounds. > > > > ... > > > > An Australian science minister previously said in January that countries needed to be "very careful" about DeepSeek, citing "data and privacy" concerns. > > > > The chatbot was removed from app stores after its privacy policy was questioned in Italy. The Italian goverment previously temporarily blocked ChatGPT over privacy concerns in March 2023. > > > > Regulators in South Korea, Ireland and France have all begun investigations into how DeepSeek handles user data, which it stores in servers in China. > > > > ... > > > > Generally, AI tools will analyse the prompts sent to them to improve their product. > > > > This is true of apps such as ChatGPT and Google Gemini as much as it is DeepSeek. > > > > All of them gather and keep information, including email addresses and dates of birth. > > > > ...

0
www.bbc.com Australia bans DeepSeek on government devices over security risk

The Australian government says the Chinese AI app is a threat to it and its assets.

Australia bans DeepSeek on government devices over security risk

cross-posted from: https://lemmy.sdf.org/post/28980041

> Australia has banned DeepSeek from all government devices and systems over what it says is the security risk the Chinese artificial intelligence (AI) startup poses. > > ... > > Growing - and familiar - concerns > > Western countries have a track record of being suspicious of Chinese tech - notably telecoms firm Huawei and the social media platform, TikTok - both of which have been restricted on national security grounds. > > ... > > An Australian science minister previously said in January that countries needed to be "very careful" about DeepSeek, citing "data and privacy" concerns. > > The chatbot was removed from app stores after its privacy policy was questioned in Italy. The Italian goverment previously temporarily blocked ChatGPT over privacy concerns in March 2023. > > Regulators in South Korea, Ireland and France have all begun investigations into how DeepSeek handles user data, which it stores in servers in China. > > ... > > Generally, AI tools will analyse the prompts sent to them to improve their product. > > This is true of apps such as ChatGPT and Google Gemini as much as it is DeepSeek. > > All of them gather and keep information, including email addresses and dates of birth. > > ...

0
www.bbc.com Australia bans DeepSeek on government devices over security risk

The Australian government says the Chinese AI app is a threat to it and its assets.

Australia bans DeepSeek on government devices over security risk

Australia has banned DeepSeek from all government devices and systems over what it says is the security risk the Chinese artificial intelligence (AI) startup poses.

...

Growing - and familiar - concerns

Western countries have a track record of being suspicious of Chinese tech - notably telecoms firm Huawei and the social media platform, TikTok - both of which have been restricted on national security grounds.

...

An Australian science minister previously said in January that countries needed to be "very careful" about DeepSeek, citing "data and privacy" concerns.

The chatbot was removed from app stores after its privacy policy was questioned in Italy. The Italian goverment previously temporarily blocked ChatGPT over privacy concerns in March 2023.

Regulators in South Korea, Ireland and France have all begun investigations into how DeepSeek handles user data, which it stores in servers in China.

...

Generally, AI tools will analyse the prompts sent to them to improve their product.

This is true of apps such as ChatGPT and Google Gemini as much as it is DeepSeek.

All of them gather and keep information, including email addresses and dates of birth.

...

18

No, DeepSeek isn’t uncensored if you run it locally

techcrunch.com No, DeepSeek isn't uncensored if you run it locally | TechCrunch

DeepSeek's model is censored at both the application and training layers, a Wired investigation shows.

No, DeepSeek isn't uncensored if you run it locally | TechCrunch

cross-posted from: https://lemmy.sdf.org/post/28978937

> There’s an idea floating around that DeepSeek’s well-documented censorship only exists at its application layer but goes away if you run it locally (that means downloading its AI model to your computer). > > But DeepSeek’s censorship is baked-in, according to a Wired investigation which found that the model is censored on both the application and training levels. > > For example, a locally run version of DeepSeek revealed to Wired thanks to its reasoning feature that it should “avoid mentioning” events like the Cultural Revolution and focus only on the “positive” aspects of the Chinese Communist Party. > > A quick check by TechCrunch of a locally run version of DeepSeek available via Groq also showed clear censorship: DeepSeek happily answered a question about the Kent State shootings in the U.S., but replied “I cannot answer” when asked about what happened in Tiananmen Square in 1989.

1

No, DeepSeek isn’t uncensored if you run it locally

techcrunch.com No, DeepSeek isn't uncensored if you run it locally | TechCrunch

DeepSeek's model is censored at both the application and training layers, a Wired investigation shows.

No, DeepSeek isn't uncensored if you run it locally | TechCrunch

cross-posted from: https://lemmy.sdf.org/post/28978937

> There’s an idea floating around that DeepSeek’s well-documented censorship only exists at its application layer but goes away if you run it locally (that means downloading its AI model to your computer). > > But DeepSeek’s censorship is baked-in, according to a Wired investigation which found that the model is censored on both the application and training levels. > > For example, a locally run version of DeepSeek revealed to Wired thanks to its reasoning feature that it should “avoid mentioning” events like the Cultural Revolution and focus only on the “positive” aspects of the Chinese Communist Party. > > A quick check by TechCrunch of a locally run version of DeepSeek available via Groq also showed clear censorship: DeepSeek happily answered a question about the Kent State shootings in the U.S., but replied “I cannot answer” when asked about what happened in Tiananmen Square in 1989.

7
Technology @midwest.social Hotznplotzn @lemmy.sdf.org

No, DeepSeek isn’t uncensored if you run it locally

techcrunch.com No, DeepSeek isn't uncensored if you run it locally | TechCrunch

DeepSeek's model is censored at both the application and training layers, a Wired investigation shows.

No, DeepSeek isn't uncensored if you run it locally | TechCrunch

There’s an idea floating around that DeepSeek’s well-documented censorship only exists at its application layer but goes away if you run it locally (that means downloading its AI model to your computer).

But DeepSeek’s censorship is baked-in, according to a Wired investigation which found that the model is censored on both the application and training levels.

For example, a locally run version of DeepSeek revealed to Wired thanks to its reasoning feature that it should “avoid mentioning” events like the Cultural Revolution and focus only on the “positive” aspects of the Chinese Communist Party.

A quick check by TechCrunch of a locally run version of DeepSeek available via Groq also showed clear censorship: DeepSeek happily answered a question about the Kent State shootings in the U.S., but replied “I cannot answer” when asked about what happened in Tiananmen Square in 1989.

8

2024 saw Europe post a record year for the production and use of renewable energies, leading the way for clean energy in the world, report says

impakter.com Europe Sets New Record For Renewable Energy Use - Impakter

2024 saw Europe post a record year for the production and use of renewable energies, leading the way for clean energy in the world.

Europe Sets New Record For Renewable Energy Use - Impakter

cross-posted from: https://lemmy.sdf.org/post/28977775

> In the European Union (EU), 47% of electricity now comes from renewable sources like wind and solar, a new record according to a report from the think tank Ember. This is a far higher percentage than in other countries, including the United States and China, where about two-thirds of energy comes from fossil fuels such as oil, coal, and gas. > > For the first time in 2024, solar energy surpassed coal in generating electricity across all 27 EU member states, while natural gas production of electricity fell for the fifth year running. > > Climate change experts are encouraged by Europe’s continued push toward cleaner energy, especially as the new U.S. administration pushes for increased fossil fuel use. > > “Fossil fuels are losing their grip on EU energy,” said Chris Rosslowe, the lead author of Ember’s ‘European Electricity Review’. “Back in 2019, at the start of the European Green Deal, few believed the energy transition would be where it is today: wind and solar are replacing coal and driving gas into a structural decline.” > > The share of electricity produced by renewables jumped to 47% last year compared to 34% in 2019, in large part due to strong growth in solar and wind energy. In 2024, 11% of the EU’s electricity came from solar power, 17% from wind, and 24% from nuclear. The share produced by traditional fossil fuels dropped from 39% in 2019 to 29% in 2024. > > ...

[Edit typo.]

0

2024 saw Europe post a record year for the production and use of renewable energies, leading the way for clean energy in the world, report says

impakter.com Europe Sets New Record For Renewable Energy Use - Impakter

2024 saw Europe post a record year for the production and use of renewable energies, leading the way for clean energy in the world.

Europe Sets New Record For Renewable Energy Use - Impakter

cross-posted from: https://lemmy.sdf.org/post/28977775

> In the European Union (EU), 47% of electricity now comes from renewable sources like wind and solar, a new record according to a report from the think tank Ember. This is a far higher percentage than in other countries, including the United States and China, where about two-thirds of energy comes from fossil fuels such as oil, coal, and gas. > > For the first time in 2024, solar energy surpassed coal in generating electricity across all 27 EU member states, while natural gas production of electricity fell for the fifth year running. > > Climate change experts are encouraged by Europe’s continued push toward cleaner energy, especially as the new U.S. administration pushes for increased fossil fuel use. > > “Fossil fuels are losing their grip on EU energy,” said Chris Rosslowe, the lead author of Ember’s ‘European Electricity Review’. “Back in 2019, at the start of the European Green Deal, few believed the energy transition would be where it is today: wind and solar are replacing coal and driving gas into a structural decline.” > > The share of electricity produced by renewables jumped to 47% last year compared to 34% in 2019, in large part due to strong growth in solar and wind energy. In 2024, 11% of the EU’s electricity came from solar power, 17% from wind, and 24% from nuclear. The share produced by traditional fossil fuels dropped from 39% in 2019 to 29% in 2024. > > ...

[Edit typo.]

0

2024 saw Europe post a record year for the production and use of renewable energies, leading the way for clean energy in the world, report says

impakter.com Europe Sets New Record For Renewable Energy Use - Impakter

2024 saw Europe post a record year for the production and use of renewable energies, leading the way for clean energy in the world.

Europe Sets New Record For Renewable Energy Use - Impakter

In the European Union (EU), 47% of electricity now comes from renewable sources like wind and solar, a new record according to a report from the think tank Ember. This is a far higher percentage than in other countries, including the United States and China, where about two-thirds of energy comes from fossil fuels such as oil, coal, and gas.

For the first time in 2024, solar energy surpassed coal in generating electricity across all 27 EU member states, while natural gas production of electricity fell for the fifth year running.

Climate change experts are encouraged by Europe’s continued push toward cleaner energy, especially as the new U.S. administration pushes for increased fossil fuel use.

“Fossil fuels are losing their grip on EU energy,” said Chris Rosslowe, the lead author of Ember’s ‘European Electricity Review’. “Back in 2019, at the start of the European Green Deal, few believed the energy transition would be where it is today: wind and solar are replacing coal and driving gas into a structural decline.”

The share of electricity produced by renewables jumped to 47% last year compared to 34% in 2019, in large part due to strong growth in solar and wind energy. In 2024, 11% of the EU’s electricity came from solar power, 17% from wind, and 24% from nuclear. The share produced by traditional fossil fuels dropped from 39% in 2019 to 29% in 2024.

...

[Edit typo.]

1
Australian economist argues China is conning the world on net zero
  • Yeah, the European Union is also good. For the first time in 2024, solar energy in the EU surpassed coal in generating electricity across all 27 EU member states, while natural gas production of electricity fell for the fifth year running.

    In the European Union (EU), 47% of electricity now comes from renewable sources like wind and solar, a new record according to a report from the think tank Ember. This is a far higher percentage than in other countries, including the United States and China, where about two-thirds of energy comes from fossil fuels such as oil, coal, and gas.

  • Australian economist argues China is conning the world on net zero
  • I don't remember when and where, but some time ago I read a blog entry where the writer used the term, "The West bad, China bad okay". Don't get me wrong, it's not about your comment alone, it's all the others in this and other threads. If China is criticized, there come all sorts of whataboutery and distractions -some of them true, some aren't- but it indeed follows this "The West bad, China bad okay" stance. The fact that China is failing on climate change is simply ignored, while every other country is criticized.

  • Australian economist argues China is conning the world on net zero
  • That's an oversimplification of the article. It's not that "we should burn fossil fuels." China should rather stop burning them.

    China's measures to fight climate change are highly insufficient as per practically all independent metrics, and the Chinese government doesn't appear to be even willing to reduce its emissions. It keeps on to produce a massive amount of overcapacity to to flood the world (and especially the global south) with cheap products for geopolitical and economic gains.

  • Australian economist argues China is conning the world on net zero

    cross-posted from: https://lemmy.sdf.org/post/28973386

    > Archived > > An Australian economist has argued China has “hoodwinked” the world on net zero in what he acknowledges could be seen as a “controversial” take. > > Leith Van Onselen, chief economist at the MB Fund and MB Super and co-founder of MacroBusiness, told 4BC host Mike Jeffreys that while Australia and other countries are making sacrifices for the good of the planet, the world’s biggest greenhouse gas emitter is not, and therefore “we’re basically shooting ourselves in the foot”. > > “I’d argue China is the largest beneficiary of the West’s net zero policies,” Mr Van Onselen said on the radio show. > > “It’s not abiding by the same rules and what we’re effectively doing is, while we’re reducing our carbon footprints through these sorts of policies and giving us expensive energy and shutting down manufacturing and becoming less diversified economies, China is simply expanding its carbon output and expanding its manufacturing sector at our expense. > > “The world’s not getting cleaner. We’re just shifting the pollution away from us, to China, and it’s actually less regulated in China.” > > ... > > He referred to a Reuters report from August that found China had boosted domestic coal production and imports to record highs. > > It said production from China’s mines hit a seasonal record of 390 million tonnes in July 2024, up from 378 million in the same month a year earlier, and imports surged to a seasonal record of 296 million tonnes in the first seven months from 261 million tonnes in 2023. > > Mr Van Onselen concluded: “I think we can all agree we’d like to have a cleaner planet but the whole idea of net zero is completely impossible or ridiculous if China is not involved. > > “If we’re going to make sacrifices, China has to make sacrifices.”

    0

    Norway to Follow EU in Suspending Visa Free Regime for Georgia’s Diplomatic Passport Holders

    As the European Union suspended visa-free travel to the Schengen area for Georgian travelers with diplomatic and official passports, “Norway will consider taking parallel steps,” the country’s Foreign Ministry said in its February 3 statement.

    The statement recalls the October 26 elections that were marred by “widespread irregularities,” then mentions the November 28 decision by the Georgian Dream government to abandon EU accession efforts, followed by non-stop protests. It highlights the authorities’ “brutal violence” against demonstrators, including at the recent February 2 protest at the entrance to Tbilisi. ...

    The EU suspended visa-free regime for Georgian diplomatic and service passport holders on January 27. This “will be followed up by parallel steps by Norway,” the statement says.

    Although the authorities can no longer enter the Schengen area without a visa with diplomatic passports, they can do so with ordinary passports, which they also possess. There have been calls for the EU not to leave this loophole to the authorities and to draw up a list of individuals who should be individually banned from entering the EU without a visa, even with ordinary passports.

    0

    Australian economist argues China is conning the world on net zero

    cross-posted from: https://lemmy.sdf.org/post/28973386

    > Archived > > An Australian economist has argued China has “hoodwinked” the world on net zero in what he acknowledges could be seen as a “controversial” take. > > Leith Van Onselen, chief economist at the MB Fund and MB Super and co-founder of MacroBusiness, told 4BC host Mike Jeffreys that while Australia and other countries are making sacrifices for the good of the planet, the world’s biggest greenhouse gas emitter is not, and therefore “we’re basically shooting ourselves in the foot”. > > “I’d argue China is the largest beneficiary of the West’s net zero policies,” Mr Van Onselen said on the radio show. > > “It’s not abiding by the same rules and what we’re effectively doing is, while we’re reducing our carbon footprints through these sorts of policies and giving us expensive energy and shutting down manufacturing and becoming less diversified economies, China is simply expanding its carbon output and expanding its manufacturing sector at our expense. > > “The world’s not getting cleaner. We’re just shifting the pollution away from us, to China, and it’s actually less regulated in China.” > > ... > > He referred to a Reuters report from August that found China had boosted domestic coal production and imports to record highs. > > It said production from China’s mines hit a seasonal record of 390 million tonnes in July 2024, up from 378 million in the same month a year earlier, and imports surged to a seasonal record of 296 million tonnes in the first seven months from 261 million tonnes in 2023. > > Mr Van Onselen concluded: “I think we can all agree we’d like to have a cleaner planet but the whole idea of net zero is completely impossible or ridiculous if China is not involved. > > “If we’re going to make sacrifices, China has to make sacrifices.”

    9

    Australian economist argues China is conning the world on net zero

    cross-posted from: https://lemmy.sdf.org/post/28973386

    > Archived > > An Australian economist has argued China has “hoodwinked” the world on net zero in what he acknowledges could be seen as a “controversial” take. > > Leith Van Onselen, chief economist at the MB Fund and MB Super and co-founder of MacroBusiness, told 4BC host Mike Jeffreys that while Australia and other countries are making sacrifices for the good of the planet, the world’s biggest greenhouse gas emitter is not, and therefore “we’re basically shooting ourselves in the foot”. > > “I’d argue China is the largest beneficiary of the West’s net zero policies,” Mr Van Onselen said on the radio show. > > “It’s not abiding by the same rules and what we’re effectively doing is, while we’re reducing our carbon footprints through these sorts of policies and giving us expensive energy and shutting down manufacturing and becoming less diversified economies, China is simply expanding its carbon output and expanding its manufacturing sector at our expense. > > “The world’s not getting cleaner. We’re just shifting the pollution away from us, to China, and it’s actually less regulated in China.” > > ... > > He referred to a Reuters report from August that found China had boosted domestic coal production and imports to record highs. > > It said production from China’s mines hit a seasonal record of 390 million tonnes in July 2024, up from 378 million in the same month a year earlier, and imports surged to a seasonal record of 296 million tonnes in the first seven months from 261 million tonnes in 2023. > > Mr Van Onselen concluded: “I think we can all agree we’d like to have a cleaner planet but the whole idea of net zero is completely impossible or ridiculous if China is not involved. > > “If we’re going to make sacrifices, China has to make sacrifices.”

    0

    Australian economist argues China is conning the world on net zero

    Archived

    An Australian economist has argued China has “hoodwinked” the world on net zero in what he acknowledges could be seen as a “controversial” take.

    Leith Van Onselen, chief economist at the MB Fund and MB Super and co-founder of MacroBusiness, told 4BC host Mike Jeffreys that while Australia and other countries are making sacrifices for the good of the planet, the world’s biggest greenhouse gas emitter is not, and therefore “we’re basically shooting ourselves in the foot”.

    “I’d argue China is the largest beneficiary of the West’s net zero policies,” Mr Van Onselen said on the radio show.

    “It’s not abiding by the same rules and what we’re effectively doing is, while we’re reducing our carbon footprints through these sorts of policies and giving us expensive energy and shutting down manufacturing and becoming less diversified economies, China is simply expanding its carbon output and expanding its manufacturing sector at our expense.

    “The world’s not getting cleaner. We’re just shifting the pollution away from us, to China, and it’s actually less regulated in China.”

    ...

    He referred to a Reuters report from August that found China had boosted domestic coal production and imports to record highs.

    It said production from China’s mines hit a seasonal record of 390 million tonnes in July 2024, up from 378 million in the same month a year earlier, and imports surged to a seasonal record of 296 million tonnes in the first seven months from 261 million tonnes in 2023.

    Mr Van Onselen concluded: “I think we can all agree we’d like to have a cleaner planet but the whole idea of net zero is completely impossible or ridiculous if China is not involved.

    “If we’re going to make sacrifices, China has to make sacrifices.”

    8

    In Bayern wird eine Klimaaktivistin nicht als Referendatin zugelassen. Ein früherer Rechtsextremist dagegen darf im Freistaat als Lehrer arbeiten.

    taz.de Debatte um Berufsverbot in Bayern: Rechts außen klappt’s mit der Schule

    In Bayern wird eine Klimaaktivistin nicht als Referendatin zugelassen. Ein früherer Rechtsextremist dagegen darf im Freistaat als Lehrer arbeiten.

    Debatte um Berufsverbot in Bayern: Rechts außen klappt’s mit der Schule

    ...

    Seit Dienstantritt [in seiner Schule] habe Michael Z. aber „in Wort und Tat nicht den geringsten Zweifel daran aufkommen lassen, dass er mit inakzeptablen Einstellungen und Aktivitäten in seiner Jugend/Studentenzeit gebrochen hat, davon gebührend Abstand hält“, so der Schulleiter.

    Michael Z. selbst antwortete auf eine taz-Anfrage nicht. Eine öffentliche Erklärung zu seinem Szeneausstieg ist nicht bekannt. Und auch die Mitgliedschaft in einer Burschenschaft gilt eigentlich auf Lebenszeit. Die „Raczek“-Burschenschaft ließ eine Anfrage unbeantwortet, ob Michael Z. bei ihnen noch Mitglied ist.

    Der Klimaaktivistin Lisa Poettinger wirft das von den Freien Wählern geführte Kultusministerium dagegen die Mitgliedschaft im „Offenen Antikapitalistischen Klimatreffen München“ vor, das vom Verfassungsschutz als extremistisch eingestuft wird.

    Verwiesen wird auch auf ihre Kapitalismuskritik bei Protesten gegen die Automesse IAA. Und auf laufende Strafverfahren wegen eines zerstörten AfD-Plakats und Widerstand gegen einen Polizisten bei der Besetzung des Dorfes Lützeraths, das für einen Kohletagebau geräumt wurde. Urteile sind hier jedoch noch nicht gefallen. Poettinger hatte angekündigt, sich juristisch gegen eine Nichtzulassung zum Referendariat wehren zu wollen.

    4

    The development of DeepSeek-V3 was probably much more expensive than suggested by the Chinese company, researchers say

    www.heise.de DeepSeek: V3 development is said to have been much more expensive

    The development of DeepSeek-V3 was probably much more expensive than suggested. The company is said to have access to 60,000 GPUs.

    DeepSeek: V3 development is said to have been much more expensive

    cross-posted from: https://lemmy.sdf.org/post/28971543

    > Archived > > DeepSeek is said to have access to tens of thousands of GPU accelerators for the development of its own AI models, including H100 GPUs, which fall under the US export bans. The reported costs of just under 5.6 million US dollars for DeepSeek v3 probably only represent a small part of the total bill. > > In the paper on the V3 model, DeepSeek writes of a comparatively small data center with 2048 H800 accelerators from Nvidia. The company calculates hypothetical rental costs of 2 US dollars per hour and H800 GPU. With a total of just under 2.8 million computing hours (distributed across 2048 GPUs), this comes to 5.6 million US dollars. > > However, the developers themselves cite a caveat: "Please note that the above costs only include the official training of DeepSeek-V3 and not the costs associated with previous research and ablation experiments on architectures, algorithms or data." > > ... > > Semianalysis has looked at a realistic cost breakdown. According to the analysts, DeepSeek has access to about 60,000 Nvidia accelerators through its parent company High-Flyer: 10,000 A100s from the Ampere generation before the US export restrictions came into effect, 10,000 H100s from the gray market, 10,000 H800s customized for China, and 30,000 H20s that Nvidia launched after more recent export restrictions. > > ... > > Semianalysis calculates that the servers required for the 60,000 GPUs cost around 1.6 billion US dollars. The operating costs are on top of that. This does not include the salaries of the development teams. > > According to DeepSeek, 96 percent of the 5.6 million US dollars quoted is for pre-training. This involves training the final underlying model. The paper ignores the previous development effort, including all the innovations incorporated into DeepSeek V2.

    0

    Die Entwicklung von DeepSeek-V3 war vermutlich viel teurer als suggeriert - Report

    www.heise.de DeepSeek soll V3 mit viel stärkerer Nvidia-Hardware entwickelt haben

    Die Entwicklung von DeepSeek-V3 war vermutlich viel teurer als suggeriert. Die Firma soll Zugriff auf 60.000 GPUs haben.

    DeepSeek soll V3 mit viel stärkerer Nvidia-Hardware entwickelt haben

    Archiv-Link

    DeepSeek soll Zugriff auf Zehntausende GPU-Beschleuniger zur Entwicklung der eigenen KI-Modelle haben, darunter auch H100-GPUs, die unter die US-Exportverbote fallen. Die kolportierten Kosten von knapp 5,6 Millionen US-Dollar für DeepSeek v3 stellen vermutlich nur einen kleinen Teil der Gesamtrechnung dar.

    Im Paper zum V3-Modell schreibt DeepSeek von einem vergleichsweise kleinen Rechenzentrum mit 2048 H800-Beschleunigern von Nvidia. Die Firma rechnet mit hypothetischen Mietkosten von 2 US-Dollar pro Stunde und H800-GPU. Bei aufsummiert knapp 2,8 Millionen Rechenstunden (verteilt auf 2048 GPUs) kommen die 5,6 Millionen US-Dollar zustande.

    Die Entwickler selbst führen allerdings eine Einschränkung an: "Bitte beachte, dass die oben genannten Kosten nur das offizielle Training von DeepSeek-V3 beinhalten und nicht die Kosten, die mit früheren Forschungs- und Ablationsexperimenten zu Architekturen, Algorithmen oder Daten verbunden sind."

    ...

    Semianalysis hat sich mit einer realistischen Kostenaufstellung beschäftigt. Laut den Analysten hat DeepSeek über das Mutterunternehmen High-Flyer Zugriff auf etwa 60.000 Nvidia-Beschleuniger: 10.000 A100 aus der Ampere-Generation, bevor die US-Exportbeschränkungen in Kraft traten, 10.000 H100 vom Graumarkt, 10.000 für China angepasste H800 und 30.000 H20, die Nvidia nach neueren Exporteinschränkungen auflegte.

    ...

    Allein die Entwicklung der Caching-Technik Multi-Head Latent Attention (MLA) soll Monate beansprucht haben. Dabei komprimiert das KI-Modell alle generierten Token, damit es bei neuen Abfragen schnell auf die Daten zugreifen kann, ohne viel Speicherplatz zu beanspruchen.

    Eine zweite Neuerung dürfte ebenfalls einige Ressourcen in Anspruch genommen haben: "Dual Pipe". DeepSeek nutzt einen Teil der Streaming-Multiprozessoren (SMs) in Nvidias GPUs als eine Art virtuelle Data Processing Unit (DPU), wie Nextplatform herausstellt. Sie kümmern sich unabhängig vom Prozessor um die Datenbewegungen in und zwischen den KI-Beschleunigern – mit viel geringeren Wartezeiten als beim Einsatz von CPUs, was die Effizienz steigert.

    Im Paper zum mächtigeren R1-Modell macht DeepSeek keinerlei Angaben zur eingesetzten Hardware. Hier wäre der Einsatz eines kleinen Rechenzentrums noch unglaubwürdiger. Ausgehend von einem X-Beitrag häuften sich zuletzt Meldungen, dass DeepSeek für R1 auch KI-Beschleuniger von Huawei verwenden könnte.

    6
    Technology @midwest.social Hotznplotzn @lemmy.sdf.org

    The development of DeepSeek-V3 was probably much more expensive than suggested by the Chinese company, researchers say

    www.heise.de DeepSeek: V3 development is said to have been much more expensive

    The development of DeepSeek-V3 was probably much more expensive than suggested. The company is said to have access to 60,000 GPUs.

    DeepSeek: V3 development is said to have been much more expensive

    cross-posted from: https://lemmy.sdf.org/post/28971543

    > Archived > > DeepSeek is said to have access to tens of thousands of GPU accelerators for the development of its own AI models, including H100 GPUs, which fall under the US export bans. The reported costs of just under 5.6 million US dollars for DeepSeek v3 probably only represent a small part of the total bill. > > In the paper on the V3 model, DeepSeek writes of a comparatively small data center with 2048 H800 accelerators from Nvidia. The company calculates hypothetical rental costs of 2 US dollars per hour and H800 GPU. With a total of just under 2.8 million computing hours (distributed across 2048 GPUs), this comes to 5.6 million US dollars. > > However, the developers themselves cite a caveat: "Please note that the above costs only include the official training of DeepSeek-V3 and not the costs associated with previous research and ablation experiments on architectures, algorithms or data." > > ... > > Semianalysis has looked at a realistic cost breakdown. According to the analysts, DeepSeek has access to about 60,000 Nvidia accelerators through its parent company High-Flyer: 10,000 A100s from the Ampere generation before the US export restrictions came into effect, 10,000 H100s from the gray market, 10,000 H800s customized for China, and 30,000 H20s that Nvidia launched after more recent export restrictions. > > ... > > Semianalysis calculates that the servers required for the 60,000 GPUs cost around 1.6 billion US dollars. The operating costs are on top of that. This does not include the salaries of the development teams. > > According to DeepSeek, 96 percent of the 5.6 million US dollars quoted is for pre-training. This involves training the final underlying model. The paper ignores the previous development effort, including all the innovations incorporated into DeepSeek V2.

    1

    The development of DeepSeek-V3 was probably much more expensive than suggested by the Chinese company, researchers say

    www.heise.de DeepSeek: V3 development is said to have been much more expensive

    The development of DeepSeek-V3 was probably much more expensive than suggested. The company is said to have access to 60,000 GPUs.

    DeepSeek: V3 development is said to have been much more expensive

    Archived

    DeepSeek is said to have access to tens of thousands of GPU accelerators for the development of its own AI models, including H100 GPUs, which fall under the US export bans. The reported costs of just under 5.6 million US dollars for DeepSeek v3 probably only represent a small part of the total bill.

    In the paper on the V3 model, DeepSeek writes of a comparatively small data center with 2048 H800 accelerators from Nvidia. The company calculates hypothetical rental costs of 2 US dollars per hour and H800 GPU. With a total of just under 2.8 million computing hours (distributed across 2048 GPUs), this comes to 5.6 million US dollars.

    However, the developers themselves cite a caveat: "Please note that the above costs only include the official training of DeepSeek-V3 and not the costs associated with previous research and ablation experiments on architectures, algorithms or data."

    ...

    Semianalysis has looked at a realistic cost breakdown. According to the analysts, DeepSeek has access to about 60,000 Nvidia accelerators through its parent company High-Flyer: 10,000 A100s from the Ampere generation before the US export restrictions came into effect, 10,000 H100s from the gray market, 10,000 H800s customized for China, and 30,000 H20s that Nvidia launched after more recent export restrictions.

    ...

    Semianalysis calculates that the servers required for the 60,000 GPUs cost around 1.6 billion US dollars. The operating costs are on top of that. This does not include the salaries of the development teams.

    According to DeepSeek, 96 percent of the 5.6 million US dollars quoted is for pre-training. This involves training the final underlying model. The paper ignores the previous development effort, including all the innovations incorporated into DeepSeek V2.

    0
    Algorithmic manipulation? TikTok use predicts positive views of China’s human rights record
  • China, in particular, has developed sophisticated strategies to control narratives and influence public opinion through digital platforms. This phenomenon, often referred to as “networked authoritarianism,” involves state actors using subtle tactics like algorithmic manipulation and strategic content curation to shape narratives on popular social media platforms.

  • Focus: DeepSeek gives Europe's tech firms a chance to catch up in global AI race
  • In related news:

    Researchers say they had a ‘100% attack success rate’ on jailbreak attempts against Chinese AI DeepSeek

    Using algorithmic jailbreaking techniques, our team applied an automated attack methodology on DeepSeek R1 which tested it against 50 random prompts from the HarmBench dataset. These covered six categories of harmful behaviors including cybercrime, misinformation, illegal activities, and general harm.

    The results were alarming: DeepSeek R1 exhibited a 100% attack success rate, meaning it failed to block a single harmful prompt. This contrasts starkly with other leading models, which demonstrated at least partial resistance.


    CNBC reports that DeepSeek’s privacy policy “isn’t worth the paper it is written on.”

    Seems to be a long way to go, but Hugging Face developers are in the process of building a fully open reproduction of DeepSeek-R1 as the AI is not Open Source as it claims.

  • Opinion | China’s gender conservations must go offline to empower everyone
  • The first point that I found suspicious with this post is the source. The SCMP is controlled by the Chinese government.

    How does that 'empower' people if they 'must go offline' to discuss whatever they want? This doesn't make sense.