{"id":15042,"date":"2025-02-01T12:28:46","date_gmt":"2025-02-01T11:28:46","guid":{"rendered":"http:\/\/plus.maciejpiasecki.info\/index.php\/2025\/02\/01\/deepseek-ais-low-cost-is-likely-a-fictional-story-expert-says\/"},"modified":"2025-02-01T21:02:23","modified_gmt":"2025-02-01T20:02:23","slug":"deepseek-ais-low-cost-is-likely-a-fictional-story-expert-says","status":"publish","type":"post","link":"https:\/\/plus.maciejpiasecki.info\/index.php\/2025\/02\/01\/deepseek-ais-low-cost-is-likely-a-fictional-story-expert-says\/","title":{"rendered":"DeepSeek AI's low cost is "likely a fictional story", expert says"},"content":{"rendered":"

DeepSeek\u2019s arrival has caused a huge stir in the artificial intelligence industry. Its AI models achieve comparable or superior performance to its competitors on multiple parameters. What\u2019s most impressive is the company\u2019s claims about the low cost invested in the training process. However, some experts question DeepSeek AI\u2019s alleged low cost, with one even saying it\u2019s \u201clikely a fictional story.\u201d
\nAnalyst says DeepSeek AI\u2019s claimed low training cost is \u201clikely a fictional story\u201c
\nAI-focused Western companies typically spend hundreds of millions of dollars on AI training alone. Next-generation advanced models are estimated to require billions in expenses even. Yet somehow, DeepSeek AI managed to train its most advanced AI model, \u201cR1,\u201d using only $6 million, or so the company claims. Not only that, but they even developed it in just a couple of months.
\nDeepSeek claims it achieved the massive cost reduction by turning to older NVIDIA H800 chips instead of more modern hardware. The collaborative approach of the Chinese AI industry was also said to have helped. However, Wedbush analyst Dan Ives labeled DeepSeek\u2019s claims as \u201clikely a fictional story.\u201d For reference, OpenAI\u2019s GPT-4 model required about $100 million in AI training only.
\nFurthermore, OpenAI suspects that DeepSeek used a \u201cshortcut\u201d in the process of training its AI. The Chinese company could have integrated data from GPT models into its own using distillation. This would have helped them save a lot of resources, but OpenAI\u2019s terms of use prohibit it. OpenAI encourages third parties to build apps on its API, but not to mine its data to train other LLMs.
\nOther experts that question DeepSeek\u2019s alleged low cost
\nThere are also questions related to the AI training \u200b\u200bhardware allegedly used by DeepSeek. Scale AI CEO Alexandr Wang said that \u201cDeepSeek has about 50,000 NVIDIA H100s that they can\u2019t talk about,\u201d as US trade restrictions prohibit them from buying them. Elon Musk backed up this claim with a simple and straightforward \u201cObviously\u201d on X\/Twitter.<\/p>\n

Obviously
\n\u2014 Elon Musk (@elonmusk) January 27, 2025<\/p>\n

Josh Kushner, founder of Thrive Capital and a major investor in OpenAI, also said that DeepSeek likely violated current US trade restrictions. Palmer Luckey, founder of Oculus, is another prominent name who raises doubts about the alleged less than $6 million invested by the Chinese firm. Plus, a US official recently claimed that DeekSeek used stolen American technology.
\nIt\u2019s noteworthy that DeepSeek has some serious financial muscle behind it. The company is backed by the Chinese High-Flyer fund which has a valuation of $8 billion. Billionaire investor Bill Ackman raised the possibility that DeepSeek AI\u2019s alleged low cost was a High-Flyer strategy to benefit from a crash in the shares of NVIDIA and others.
\nThe post DeepSeek AI’s low cost is „likely a fictional story”, expert says appeared first on Android Headlines.
\n
\nSource: ndroidheadlines.com <\/p>\n","protected":false},"excerpt":{"rendered":"

DeepSeek\u2019s arrival has caused a huge stir in the artificial intelligence industry. Its AI models achieve comparable or superior performance […]<\/p>\n","protected":false},"author":67,"featured_media":15043,"comment_status":"false","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-15042","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-bez-kategorii"],"_links":{"self":[{"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/posts\/15042","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/users\/67"}],"replies":[{"embeddable":true,"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/comments?post=15042"}],"version-history":[{"count":1,"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/posts\/15042\/revisions"}],"predecessor-version":[{"id":15044,"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/posts\/15042\/revisions\/15044"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/media\/15043"}],"wp:attachment":[{"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/media?parent=15042"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/categories?post=15042"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/plus.maciejpiasecki.info\/index.php\/wp-json\/wp\/v2\/tags?post=15042"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}