Semantics Relevance

Microsoft Makes a New Push Into Smaller A.I. Systems


Logical Analysis Report (click to view);
Knowledge Map (click to view)
*Knowledge Map Navigation: Spatial co-ordinations are initially random, and will automatically re-arrange to minimize complexity based on distance between relationships. Mouse down and drag to pan. Right click on the strategic diagram toggles between motion and stationary. Hover over abstract node (orange) to view abstractions. Hover over leaf node to view corresponding narrative. Left click on the leaf node expands the narrative to view full text.

Knowledge Diagram Navigation:

Spatial co-ordinations are initially random, and will automatically re-arrange to minimize complexity based on distance between relationships. Mouse down and drag to pan. Right click on the strategic diagram toggles between motion and stationary. Hover over abstract node (orange) to view abstractions. Hover over leaf node to view corresponding narrative. Left click on the leaf node expands the narrative to view full text.

Narrative Analysis - Report

Key Focus

  • Building these models are a trade-off between power and size. S bastien Bubeck, a researcher and vice president at Microsoft, said the company built its new smaller models by refining the data that was pumped into them, working to ensure that the models learned from higher-quality text ...
  • Momentum supporting factors

  • (openai,start-ups)
  • Challenge supporting factors

  • (microsoft,openai)
  • (models,microsoft,openai)
  • (models,microsoft,new_york_times)
  • (models,openai,new_york_times)
  • (openai,new_york_times)
  • (microsoft,infringement)
  • (models,microsoft,infringement)
  • (models,openai,infringement)
  • (models,openai,copyright)
  • (microsoft,technologies)
  • (microsoft,new_york_times)
  • Work-in-progress supporting factors

  • (models,company,researcher)
  • (models,microsoft,researcher)
  • (models,researcher)
  • (models,company,president)
  • (models,microsoft,president)
  • (models,company,higher-quality)
  • (models,microsoft,higher-quality)
  • (models,microsoft,executive)
  • (models,company,technology)
  • (models,technology)
  • (models,company,phi-3)
  • (models,company,openai)
  • (models,company,microsoft)
  • (models,openai,gpt-3)
  • (openai,company,gpt-3)

  • Time PeriodChallengeMomentumWIP
    Report18.29 1.22 80.49

    High Level Abstraction (HLA) combined

    High Level Abstraction (HLA)Report
    (1) (models,company,researcher)100.00
    (2) (models,microsoft,researcher)94.69
    (3) (models,researcher)92.92
    (4) (models,company,president)90.27
    (5) (models,microsoft,president)85.84
    (6) (microsoft,openai)84.07
    (7) (models,microsoft,openai)82.30
    (8) (models,microsoft,new_york_times)79.65
    (9) (models,openai,new_york_times)75.22
    (10) (openai,new_york_times)73.45
    (11) (microsoft,infringement)72.57
    (12) (models,microsoft,infringement)69.91
    (13) (models,openai,infringement)65.49
    (14) (models,company,higher-quality)64.60
    (15) (models,microsoft,higher-quality)63.72
    (16) (models,microsoft,executive)59.29
    (17) (models,company,technology)51.33
    (18) (models,technology)49.56
    (19) (models,company,phi-3)47.79
    (20) (models,company,openai)44.25
    (21) (models,company,microsoft)39.82
    (22) (models,openai,gpt-3)35.40
    (23) (openai,company,gpt-3)34.51
    (24) (models,openai,copyright)33.63
    (25) (models,openai,chatbot)32.74
    (26) (models,wikipedia)31.86
    (27) (models,user)30.97
    (28) (models,situations)30.09
    (29) (microsoft,technologies)29.20
    (30) (microsoft,software)28.32
    (31) (microsoft,new_york_times)27.43
    (32) (microsoft,industry)26.55
    (33) (microsoft,gpt-4)25.66
    (34) (openai,company,wikipedia)22.12
    (35) (openai,wikipedia)20.35
    (36) (openai,company,web)17.70
    (37) (openai,web)15.04
    (38) (openai,company,services)13.27
    (39) (openai,services)9.73
    (40) (openai,company,pioneers)8.85
    (41) (openai,company,executive)4.42
    (42) (openai,company,cost)3.54
    (43) (openai,start-ups)2.65
    (44) (openai,retrieving)1.77
    (45) (openai,pioneers)0.88

    Back to top of page

    Supporting narratives:

    Please refer to knowledge diagram for a complete set of supporting narratives.

    • momentum - Back to HLA
      • But L.L.M.s store so much information, retrieving what is needed for each chat requires considerable computing power. And that is expensive.. . While tech giants and start-ups like OpenAI and Anthropic have been focused on improving the largest A.I ...
      • High Level Abstractions:
        • (openai,start-ups)

    • challenge - Back to HLA
      • Microsoft is open sourcing its new Phi-3 models, too.. . (The New York Times sued OpenAI and Microsoft in December for copyright infringement of news content related to A.I ...
      • High Level Abstractions:
        • (microsoft,openai)
        • (models,microsoft,new_york_times)
        • (models,openai,new_york_times)
        • (openai,new_york_times)
        • (models,microsoft,openai)
        • Inferred entity relationships (6)
        • (microsoft,models,new_york_times) [inferred]
        • (new_york_times,openai) [inferred]
        • (models,new_york_times,openai) [inferred]
        • (microsoft,models,researcher) [inferred]
        • (microsoft,models,president) [inferred]
        • (microsoft,models,openai) [inferred]

    • challenge - Back to HLA
      • . (The New York Times sued OpenAI and Microsoft in December for copyright infringement of news content related to A.I. ...
      • High Level Abstractions:
        • (models,openai,infringement)
        • (models,microsoft,infringement)
        • (microsoft,infringement)
        • Inferred entity relationships (6)
        • (microsoft,models,new_york_times) [inferred]
        • (infringement,microsoft,models) [inferred]
        • (infringement,microsoft) [inferred]
        • (microsoft,models,researcher) [inferred]
        • (microsoft,models,president) [inferred]
        • (microsoft,models,openai) [inferred]

    • challenge - Back to HLA
      • Microsoft is open sourcing its new Phi-3 models, too.. . (The New York Times sued OpenAI and Microsoft in December for copyright infringement of news content related to A.I. ...
      • High Level Abstractions:
        • (models,openai,copyright)

    • challenge - Back to HLA
      • This is a common way for companies to get outside help improving their software and to encourage the larger industry to use their technologies. Microsoft is open sourcing its new Phi-3 models, too.. . (The New York Times sued OpenAI and Microsoft in December for copyright infringement of news content related to A.I ...
      • High Level Abstractions:
        • (microsoft,technologies)
        • (microsoft,new_york_times)

    • WIP - Back to HLA
      • Building these models are a trade-off between power and size. S bastien Bubeck, a researcher and vice president at Microsoft, said the company built its new smaller models by refining the data that was pumped into them, working to ensure that the models learned from higher-quality text ...
      • High Level Abstractions:
        • (models,company,microsoft)
        • (models,researcher)
        • (models,microsoft,president)
        • (models,company,president)
        • (models,company,researcher)
        • (models,microsoft,researcher)
        • Inferred entity relationships (10)
        • (company,models,president) [inferred]
        • (company,models,researcher) [inferred]
        • (microsoft,models,researcher) [inferred]
        • (microsoft,models,president) [inferred]
        • (microsoft,models,openai) [inferred]
        • (microsoft,models,new_york_times) [inferred]
        • (models,researcher) [inferred]
        • (company,models,openai) [inferred]
        • (company,models,technology) [inferred]
        • (company,models,phi-3) [inferred]

    • WIP - Back to HLA
      • S bastien Bubeck, a researcher and vice president at Microsoft, said the company built its new smaller models by refining the data that was pumped into them, working to ensure that the models learned from higher-quality text. ...
      • High Level Abstractions:
        • (models,microsoft,higher-quality)
        • (models,company,higher-quality)
        • Inferred entity relationships (4)
        • (microsoft,models,new_york_times) [inferred]
        • (microsoft,models,researcher) [inferred]
        • (microsoft,models,president) [inferred]
        • (microsoft,models,openai) [inferred]

    • WIP - Back to HLA
      • Customers imagine many ways to use A.I., but with the biggest systems "they're like, 'Oh, but you know, they can get kind of expensive,'" said Eric Boyd, a Microsoft executive. Smaller models, almost by definition, are cheaper to deploy, he said ...
      • High Level Abstractions:
        • (models,microsoft,executive)
        • Inferred entity relationships (4)
        • (microsoft,models,new_york_times) [inferred]
        • (microsoft,models,researcher) [inferred]
        • (microsoft,models,president) [inferred]
        • (microsoft,models,openai) [inferred]

    • WIP - Back to HLA
      • On Tuesday, Microsoft introduced three smaller A.I. models that are part of a technology family the company has named Phi-3. The company said even the smallest of the three performed almost as well as GPT-3.5, the much larger system that underpinned OpenAI's ChatGPT chatbot when it stunned the world upon its release in late 2022 ...
      • High Level Abstractions:
        • (models,technology)
        • (models,company,technology)
        • (models,company,phi-3)
        • Inferred entity relationships (6)
        • (models,technology) [inferred]
        • (company,models,president) [inferred]
        • (company,models,researcher) [inferred]
        • (company,models,openai) [inferred]
        • (company,models,technology) [inferred]
        • (company,models,phi-3) [inferred]

    • WIP - Back to HLA
      • The company said even the smallest of the three performed almost as well as GPT-3.5, the much larger system that underpinned OpenAI's ChatGPT chatbot when it stunned the world upon its release in late 2022 ...
      • High Level Abstractions:
        • (models,openai,gpt-3)
        • (models,company,openai)
        • (openai,company,gpt-3)
        • Inferred entity relationships (4)
        • (company,models,president) [inferred]
        • (company,models,researcher) [inferred]
        • (company,models,technology) [inferred]
        • (company,models,phi-3) [inferred]

    • WIP - Back to HLA
      • The company said even the smallest of the three performed almost as well as GPT-3.5, the much larger system that underpinned OpenAI's ChatGPT chatbot when it stunned the world upon its release in late 2022. ...
      • High Level Abstractions:
        • (models,openai,chatbot)

    • WIP - Back to HLA
      • Chatbots are driven by large language models, or L.L.M.s, mathematical systems that spend weeks analyzing digital books, Wikipedia articles, news articles, chat logs and other text culled from across the internet ...
      • High Level Abstractions:
        • (models,wikipedia)

    • WIP - Back to HLA
      • "Other situations, where I am summarizing online user reviews, if it's a little bit off, it's not the end of the world." ...
      • High Level Abstractions:
        • (models,situations)
        • (models,user)

    • WIP - Back to HLA
      • This is a common way for companies to get outside help improving their software and to encourage the larger industry to use their technologies. Microsoft is open sourcing its new Phi-3 models, too. ...
      • High Level Abstractions:
        • (microsoft,industry)
        • (microsoft,software)

    • WIP - Back to HLA
      • Though Microsoft said using the new models would be "substantially cheaper" than using larger models like GPT-4, it did not offer specifics. ...
      • High Level Abstractions:
        • (microsoft,gpt-4)

    • WIP - Back to HLA
      • After OpenAI released ChatGPT, Sam Altman, the company's chief executive, said the cost of each chat was "single-digits cents" - an enormous expense considering what popular web services like Wikipedia are serving up for tiny fractions of a cent. ...
      • High Level Abstractions:
        • (openai,company,services)
        • (openai,company,web)
        • (openai,wikipedia)
        • (openai,services)
        • (openai,company,wikipedia)
        • (openai,web)
        • Inferred entity relationships (7)
        • (openai,web) [inferred]
        • (company,openai,services) [inferred]
        • (company,openai,wikipedia) [inferred]
        • (company,openai,web) [inferred]
        • (company,openai,pioneers) [inferred]
        • (openai,wikipedia) [inferred]
        • (openai,services) [inferred]

    • WIP - Back to HLA
      • . The company that has invested billions in generative A.I. pioneers like OpenAI says giant systems aren't necessarily what everyone needs ...
      • High Level Abstractions:
        • (openai,company,pioneers)
        • Inferred entity relationships (4)
        • (openai,pioneers) [inferred]
        • (company,openai,services) [inferred]
        • (company,openai,wikipedia) [inferred]
        • (company,openai,web) [inferred]

    • WIP - Back to HLA
      • systems.). . After OpenAI released ChatGPT, Sam Altman, the company's chief executive, said the cost of each chat was "single-digits cents" - an enormous expense considering what popular web services like Wikipedia are serving up for tiny fractions of a cent ...
      • High Level Abstractions:
        • (openai,company,executive)

    • WIP - Back to HLA
      • After OpenAI released ChatGPT, Sam Altman, the company's chief executive, said the cost of each chat was "single-digits cents" - an enormous expense considering what popular web services like Wikipedia are serving up for tiny fractions of a cent ...
      • High Level Abstractions:
        • (openai,company,cost)

    • WIP - Back to HLA
      • But L.L.M.s store so much information, retrieving what is needed for each chat requires considerable computing power ...
      • High Level Abstractions:
        • (openai,retrieving)

    • WIP - Back to HLA
      • . The company that has invested billions in generative A.I. pioneers like OpenAI says giant systems aren't necessarily what everyone needs. ...
      • High Level Abstractions:
        • (openai,pioneers)