Introduction
intha பதிவுல
la, namma deepseek apdingara oru popular opensource model pathi thrinjikaporom...
kurippa, Deepseek oda reasoning LLM pathi therinjikaporom
Athoda, athu yen ippo rombha populara pesa paduthun therinjikaporom...
mothella, ithoda popularity ku varuvom... Summar 1 vaarum munnadi athavathu Jan 20, 2025 anniku, deepseek ngaara oru chinese research lab, avaloga reasoning model aana deepseek-r1 ngara model huh veliyitaanga... Incentivizing reasoning capability in LLMs via Reinforcement learning ngara paper la, intha model oda perforamnce huh pala frontier properity models oda oopittu ori benchmark-aiyum veiliyittanga... Ithula mukkiyamana vishayam enna na, OpenAI oda reasoning models oda perforamnce huh verum 10 Million dollor cost la panna mudiyum nu avangua veliyitta technical report la therivichu irunthaanga... athavathu OPenaiya kaatilum sumaar 27x cheaperaana model huh koduthuirukaanga...
Apram enna.. Tech ulagathula memes enna panjama... openai va varuthu eduthutaanga...
Ithu ennamaariyana timing na... kitathatta athe vaaram, US president trump stargate nu oru puthu project huh openai oda sam altman, softbank oda masayoshi son & oracle oda larry ellisson kooda sernthu... sumaar 500B USD muthaleetla announce panni irunthaaru... ithu ethukuna... openai maathirayana companies naraya massive datacenters huh AI research kaaga US la build pannrathukaana oru oppurunity huh itha project pannirunthaanga...
aana cut pannina... inga verum 200 engineers konda oru chinese startup, oru fraction of cost la, entha vithamana periya datacenter investmnets illama... OpenAI models ku nigarana.. yen innum sollpona oru sila benchmarks la proproetary models sa minjara alavaukkana models huh create panna mudiyum nu nirubitchirukaanga... athu mattum illama, yaan petra inbam peruga ivaaiyagam nu avangola athana innovationa-aiyum open source la verliyittu.. thaaru maaru kaati-irukaanga...
NVIDIA Case (தங்க பொதையல் இருக்கற ஊர்ல, கடப்பறைய வித்தவங்களோட கதை)
AI boom arambitcha kalathula irunthu kodi katti parantha ore company na athu namma nvidia than... Ellarum thangatha thondrappa... ellaruom kadapariya supply pandra veliya partha nvidia oda valarchi sumaary 2500% .. inime namma than nu iruntha nvidia-kku.. ippo oru sarukkal nu thaan sollanum.. S&P 500 index apdi thooki piduchi iruntha nvidia kitathatta 17% correct aagapoga... athaye nambi iruntha S&P500 oda gathiyum athe than... inaikku market oda heat map evolo sooda sennirthula irukkunu neenga screen la paakalam...
DeepSeek Technical Report - Deep dive 🐋
Ithellam orupakkam irukattum.... namma oru tech channel... namma yen stock market pathi pesanam... hence athukku bathila namma deepseek R1 technical report huh rombha eliya muraila purinjika try pannuvom vaanga...
Language vs Reasoning Models
Sari mothella deepseek ngara website ku poneenga anga naraiya models irukartha neenga paakalam... deepseek v1, v2, v3 nu irukarathellam langauge models.. ithu gpt, llama, mistral mathiriyana ellorum therinja chat models. Athukku aprama deepseek R0, R1, R1 distilled nu naraya reasoning models (athavathu inga irukara intha R ngara letter) intha reasoning capability huh than kurikuthu...
Sari... intha reasoning models na enna? vaanga purinjikalam...
Keep reading with a 7-day free trial
Subscribe to Concepts in Tamil (தமிழ்) to keep reading this post and get 7 days of free access to the full post archives.