Jump to content

Gemini (language model)

From Wikipedia, the free encyclopedia
(Redirected from Gemini Nano)

Gemini
Developer(s)Google DeepMind
Initial releaseDecember 6, 2023; 12 months ago (2023-12-06)
PredecessorPaLM 2
Available inEnglish
TypeLarge language model
LicenseProprietary
Websitedeepmind.google/technologies/gemini/ Edit this at Wikidata

Gemini (formerly known as Bard) is a family of multimodal large language models developed by Google DeepMind, serving as the successor to LaMDA and PaLM 2. Comprising Gemini Ultra, Gemini Pro, Gemini Flash, and Gemini Nano, it was announced on December 6, 2023, positioned as a competitor to OpenAI's GPT-4. It powers the chatbot of the same name.

History

[edit]

Development

[edit]
Further information: Gemini (chatbot) § Background
Google CEO Sundar Pichai (L) and DeepMind CEO Demis Hassabis (R) spearheaded the development of Gemini.

Google announced Gemini, a large language model (LLM) developed by subsidiary Google DeepMind, during the Google I/O keynote on May 10, 2023. It was positioned as a more powerful successor to PaLM 2, which was also unveiled at the event, with Google CEO Sundar Pichai stating that Gemini was still in its early developmental stages.[1][2] Unlike other LLMs, Gemini was said to be unique in that it was not trained on a text corpus alone and was designed to be multimodal, meaning it could process multiple types of data simultaneously, including text, images, audio, video, and computer code.[3] It had been developed as a collaboration between DeepMind and Google Brain, two branches of Google that had been merged as Google DeepMind the previous month.[4] In an interview with Wired, DeepMind CEO Demis Hassabis touted Gemini's advanced capabilities, which he believed would allow the algorithm to trump OpenAI's ChatGPT, which runs on GPT-4 and whose growing popularity had been aggressively challenged by Google with LaMDA and Bard. Hassabis highlighted the strengths of DeepMind's AlphaGo program, which gained worldwide attention in 2016 when it defeated Go champion Lee Sedol, saying that Gemini would combine the power of AlphaGo and other Google–DeepMind LLMs.[5]

In August 2023, The Information published a report outlining Google's roadmap for Gemini, revealing that the company was targeting a launch date of late 2023. According to the report, Google hoped to surpass OpenAI and other competitors by combining conversational text capabilities present in most LLMs with artificial intelligence–powered image generation, allowing it to create contextual images and be adapted for a wider range of use cases.[6] Like Bard,[7] Google co-founder Sergey Brin was summoned out of retirement to assist in the development of Gemini, along with hundreds of other engineers from Google Brain and DeepMind;[6][8] he was later credited as a "core contributor" to Gemini.[9] Because Gemini was being trained on transcripts of YouTube videos, lawyers were brought in to filter out any potentially copyrighted materials.[6]

With news of Gemini's impending launch, OpenAI hastened its work on integrating GPT-4 with multimodal features similar to those of Gemini.[10] The Information reported in September that several companies had been granted early access to "an early version" of the LLM, which Google intended to make available to clients through Google Cloud's Vertex AI service. The publication also stated that Google was arming Gemini to compete with both GPT-4 and Microsoft's GitHub Copilot.[11][12]

Launch

[edit]

On December 6, 2023, Pichai and Hassabis announced "Gemini 1.0" at a virtual press conference.[13][14] It comprised three models: Gemini Ultra, designed for "highly complex tasks"; Gemini Pro, designed for "a wide range of tasks"; and Gemini Nano, designed for "on-device tasks". At launch, Gemini Pro and Nano were integrated into Bard and the Pixel 8 Pro smartphone, respectively, while Gemini Ultra was set to power "Bard Advanced" and become available to software developers in early 2024. Other products that Google intended to incorporate Gemini into included Search, Ads, Chrome, Duet AI on Google Workspace, and AlphaCode 2.[15][14] It was made available only in English.[14][16] Touted as Google's "largest and most capable AI model" and designed to emulate human behavior,[17][14][18] the company stated that Gemini would not be made widely available until the following year due to the need for "extensive safety testing".[13] Gemini was trained on and powered by Google's Tensor Processing Units (TPUs),[13][16] and the name is in reference to the DeepMind–Google Brain merger as well as NASA's Project Gemini.[19]

Gemini Ultra was said to have outperformed GPT-4, Anthropic's Claude 2, Inflection AI's Inflection-2, Meta's LLaMA 2, and xAI's Grok 1 on a variety of industry benchmarks,[20][13] while Gemini Pro was said to have outperformed GPT-3.5.[3] Gemini Ultra was also the first language model to outperform human experts on the 57-subject Massive Multitask Language Understanding (MMLU) test, obtaining a score of 90%.[3][19] Gemini Pro was made available to Google Cloud customers on AI Studio and Vertex AI on December 13, while Gemini Nano will be made available to Android developers as well.[21][22][23] Hassabis further revealed that DeepMind was exploring how Gemini could be "combined with robotics to physically interact with the world".[24] In accordance with an executive order signed by U.S. President Joe Biden in October, Google stated that it would share testing results of Gemini Ultra with the federal government of the United States. Similarly, the company was engaged in discussions with the government of the United Kingdom to comply with the principles laid out at the AI Safety Summit at Bletchley Park in November.[3]

Updates

[edit]

Google partnered with Samsung to integrate Gemini Nano and Gemini Pro into its Galaxy S24 smartphone lineup in January 2024.[25][26] The following month, Bard and Duet AI were unified under the Gemini brand,[27][28] with "Gemini Advanced with Ultra 1.0" debuting via a new "AI Premium" tier of the Google One subscription service.[29] Gemini Pro also received a global launch.[30]

In February, 2024, Google launched "Gemini 1.5" in a limited capacity, positioned as a more powerful and capable model than 1.0 Ultra.[31][32][33] This "step change" was achieved through various technical advancements, including a new architecture, a mixture-of-experts approach, and a larger one-million-token context window, which equates to roughly an hour of silent video, 11 hours of audio, 30,000 lines of code, or 700,000 words.[34] The same month, Google debuted Gemma, a family of free and open-source LLMs that serve as a lightweight version of Gemini. They come in two sizes, with a neural network with two and seven billion parameters, respectively. Multiple publications viewed this as a response to Meta and others open-sourcing their AI models, and a stark reversal from Google's longstanding practice of keeping its AI proprietary.[35][36][37] Google announced an additional model, Gemini 1.5 Flash, on May 14th at the 2024 I/O keynote.[38]

Gemma 2 was released on June 27, 2024.[39]

Two updated Gemini models, Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002, were released on September 24, 2024.[40]

On December 11, 2024, Google announced Gemini 2.0 Flash Experimental[41], a significant update to its Gemini AI model. This iteration boasts improved speed and performance over its predecessor, Gemini 1.5 Flash. Key features include a Multimodal Live API for real-time audio and video interactions, enhanced spatial understanding, native image and controllable text-to-speech generation (with watermarking), and integrated tool use, including Google Search.[42] It also introduces improved agentic capabilities, a new Google Gen AI SDK[43], and "Jules," an experimental AI coding agent for GitHub. Additionally, Google Colab is integrating Gemini 2.0 to generate data science notebooks from natural language. Gemini 2.0 is available through the Gemini chat interface for all users as "Gemini 2.0 Flash experimental".  Gemini 2.0 Flash is currently in an experimental phase, with a wider release expected in early 2025.

Technical specifications

[edit]

The first generation of Gemini ("Gemini 1") has three models, with the same architecture. They are decoder-only transformers, with modifications to allow efficient training and inference on TPUs. They have a context length of 32,768 tokens, with multi-query attention. Two versions of Gemini Nano, Nano-1 (1.8 billion parameters) and Nano-2 (3.25 billion parameters), are distilled from larger Gemini models, designed for use by edge devices such as smartphones. As Gemini is multimodal, each context window can contain multiple forms of input. The different modes can be interleaved and do not have to be presented in a fixed order, allowing for a multimodal conversation. For example, the user might open the conversation with a mix of text, picture, video, and audio, presented in any order, and Gemini might reply with the same free ordering. Input images may be of different resolutions, while video is inputted as a sequence of images. Audio is sampled at 16 kHz and then converted into a sequence of tokens by the Universal Speech Model. Gemini's dataset is multimodal and multilingual, consisting of "web documents, books, and code, and includ[ing] image, audio, and video data".[44]

The second generation of Gemini ("Gemini 1.5") has two models. Gemini 1.5 Pro is a multimodal sparse mixture-of-experts, with a context length in the millions, while Gemini 1.5 Flash is distilled from Gemini 1.5 Pro, with a context length above 2 million.[45]

Gemma 2 27B is trained on web documents, code, science articles. Gemma 2 9B was distilled from 27B. Gemma 2 2B was distilled from a 7B model that remained unreleased.[46]

As of 2024 August, the models released include[47]

  • Gemma 1 (2B, 7B)
  • CodeGemma (2B and 7B) - Gemma 1 finetuned for code generation.
  • Gemma 2 (2B, 9B, 27B) - 27B trained from scratch. 2B and 9B
  • RecurrentGemma (2B, 9B) - Griffin-based, instead of Transformer-based.
  • PaliGemma (3B) - A vision-language model that takes text and image inputs, and outputs text. It is based on PaLI.[48]

Reception

[edit]
See also: Gemini (chatbot) § Reception

Gemini's launch was preluded by months of intense speculation and anticipation, which MIT Technology Review described as "peak AI hype".[49][20] In August 2023, Dylan Patel and Daniel Nishball of research firm SemiAnalysis penned a blog post declaring that the release of Gemini would "eat the world" and outclass GPT-4, prompting OpenAI CEO Sam Altman to ridicule the duo on X (formerly Twitter).[50][51] Business magnate Elon Musk, who co-founded OpenAI, weighed in, asking, "Are the numbers wrong?"[52] Hugh Langley of Business Insider remarked that Gemini would be a make-or-break moment for Google, writing: "If Gemini dazzles, it will help Google change the narrative that it was blindsided by Microsoft and OpenAI. If it disappoints, it will embolden critics who say Google has fallen behind."[53]

Reacting to its unveiling in December 2023, University of Washington professor emeritus Oren Etzioni predicted a "tit-for-tat arms race" between Google and OpenAI. Professor Alexei Efros of the University of California, Berkeley praised the potential of Gemini's multimodal approach,[19] while scientist Melanie Mitchell of the Santa Fe Institute called Gemini "very sophisticated". Professor Chirag Shah of the University of Washington was less impressed, likening Gemini's launch to the routineness of Apple's annual introduction of a new iPhone. Similarly, Stanford University's Percy Liang, the University of Washington's Emily Bender, and the University of Galway's Michael Madden cautioned that it was difficult to interpret benchmark scores without insight into the training data used.[49][54] Writing for Fast Company, Mark Sullivan opined that Google had the opportunity to challenge the iPhone's dominant market share, believing that Apple was unlikely to have the capacity to develop functionality similar to Gemini with its Siri virtual assistant.[55] Google shares spiked by 5.3 percent the day after Gemini's launch.[56][57]

Google faced criticism for a demonstrative video of Gemini, which was not conducted in real time.[58]

See also

[edit]
  • Gato, a multimodal neural network developed by DeepMind

References

[edit]
  1. ^ Grant, Nico (May 10, 2023). "Google Builds on Tech's Latest Craze With Its Own A.I. Products". The New York Times. ISSN 0362-4331. Archived from the original on May 10, 2023. Retrieved August 21, 2023.
  2. ^ Ortiz, Sabrina (May 10, 2023). "Every major AI feature announced at Google I/O 2023". ZDNet. Archived from the original on May 10, 2023. Retrieved August 21, 2023.
  3. ^ a b c d Milmo, Dan (December 6, 2023). "Google says new AI model Gemini outperforms ChatGPT in most tests". The Guardian. ISSN 0261-3077. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  4. ^ Levy, Steven (September 11, 2023). "Sundar Pichai on Google;s AI, Microsoft's AI, OpenAI, and ... Did We Mention AI?". Wired. Archived from the original on September 11, 2023. Retrieved September 12, 2023.
  5. ^ Knight, Will (June 26, 2023). "Google DeepMind's CEO Says Its Next Algorithm Will Eclipse ChatGPT". Wired. Archived from the original on June 26, 2023. Retrieved August 21, 2023.
  6. ^ a b c Victor, Jon (August 15, 2023). "How Google is Planning to Beat OpenAI". The Information. Archived from the original on August 15, 2023. Retrieved August 21, 2023.
  7. ^ Grant, Nico (January 20, 2023). "Google Calls In Help From Larry Page and Sergey Brin for A.I. Fight". The New York Times. ISSN 0362-4331. Archived from the original on January 20, 2023. Retrieved February 6, 2023.
  8. ^ Kruppa, Miles; Seetharaman, Deepa (July 21, 2023). "Sergey Brin Is Back in the Trenches at Google". The Wall Street Journal. ISSN 0099-9660. Archived from the original on July 21, 2023. Retrieved September 7, 2023.
  9. ^ Carter, Tom (December 7, 2023). "Google confirms that its cofounder Sergey Brin played a key role in creating its ChatGPT rival". Business Insider. Archived from the original on December 7, 2023. Retrieved December 31, 2023.
  10. ^ Victor, Jon (September 18, 2023). "OpenAI Hustles to Beat Google to Launch 'Multimodal' LLM". The Information. Archived from the original on September 18, 2023. Retrieved October 15, 2023.
  11. ^ "Google nears release of AI software Gemini, The Information reports". Reuters. September 14, 2023. Archived from the original on September 15, 2023. Retrieved October 2, 2023.
  12. ^ Nolan, Beatrice (September 23, 2023). "Google is quietly handing out early demos of its GPT-4 rival called Gemini. Here's what we know so far about the upcoming AI model". Business Insider. Archived from the original on September 23, 2023. Retrieved October 16, 2023.
  13. ^ a b c d Kruppa, Miles (December 6, 2023). "Google Announces AI System Gemini After Turmoil at Rival OpenAI". The Wall Street Journal. ISSN 0099-9660. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  14. ^ a b c d Liedtike, Michael; O'Brien, Matt (December 6, 2023). "Google launches Gemini, upping the stakes in the global AI race". Associated Press. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  15. ^ Edwards, Benj (December 6, 2023). "Google launches Gemini—a powerful AI model it says can surpass GPT-4". Ars Technica. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  16. ^ a b Pierce, David (December 6, 2023). "Google launches Gemini, the AI model it hopes will take down GPT-4". The Verge. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  17. ^ Fung, Brian; Thorbecke, Catherine (December 6, 2023). "Google launches Gemini, its most-advanced AI model yet, as it races to compete with ChatGPT". CNN Business. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  18. ^ "Google launches Gemini, upping the stakes in the global AI race". CBS News. December 6, 2023. Archived from the original on December 7, 2023. Retrieved December 7, 2023.
  19. ^ a b c Knight, Will (December 6, 2023). "Google Just Launched Gemini, Its Long-Awaited Answer to ChatGPT". Wired. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  20. ^ a b Henshall, Will (December 6, 2023). "Google DeepMind Unveils Its Most Powerful AI Offering Yet". Time. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  21. ^ Metz, Cade; Grant, Nico (December 6, 2023). "Google Updates Bard Chatbot With 'Gemini' A.I. as It Chases ChatGPT". The New York Times. ISSN 0362-4331. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  22. ^ Elias, Jennifer (December 6, 2023). "Google launches its largest and 'most capable' AI model, Gemini". CNBC. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  23. ^ Alba, Davey; Ghaffary, Shirin (December 6, 2023). "Google Opens Access to Gemini, Racing to Catch Up to OpenAI". Bloomberg News. Archived from the original on December 6, 2023. Retrieved December 7, 2023.
  24. ^ Knight, Will (December 6, 2023). "Google DeepMind's Demis Hassabis Says Gemini Is a New Breed of AI". Wired. Archived from the original on December 6, 2023. Retrieved December 7, 2023.
  25. ^ Gurman, Mark; Love, Julia; Alba, Davey (January 17, 2024). "Samsung Bets on Google-Powered AI Features in Smartphone Revamp". Bloomberg News. Archived from the original on January 17, 2024. Retrieved February 4, 2024.
  26. ^ Chokkattu, Julian (January 17, 2024). "Samsung's Galaxy S24 Phones Call on Google's AI to Spruce Up Their Smarts". Wired. Archived from the original on January 17, 2024. Retrieved February 4, 2024.
  27. ^ Metz, Cade (February 8, 2024). "Google Releases Gemini, an A.I.-Driven Chatbot and Voice Assistant". The New York Times. ISSN 0362-4331. Archived from the original on February 8, 2024. Retrieved February 8, 2024.
  28. ^ Dastin, Jeffrey (February 8, 2024). "Google rebrands Bard chatbot as Gemini, rolls out paid subscription". Reuters. Archived from the original on February 8, 2024. Retrieved February 8, 2024.
  29. ^ Li, Abner (February 8, 2024). "Google One AI Premium is $19.99/mo with Gemini Advanced & Gemini for Workspace". 9to5Google. Archived from the original on February 8, 2024. Retrieved February 8, 2024.
  30. ^ Mehta, Ivan (February 1, 2024). "Google's Bard chatbot gets the Gemini Pro update globally". TechCrunch. Archived from the original on February 1, 2024. Retrieved February 4, 2024.
  31. ^ Knight, Will (February 15, 2024). "Google's Flagship AI Model Gets a Mighty Fast Upgrade". Wired. Archived from the original on February 15, 2024. Retrieved February 21, 2024.
  32. ^ Nieva, Richard (February 15, 2024). "Google Unveils Gemini 1.5, But Only Developers And Enterprise Clients Have Access For Now". Forbes. Archived from the original on February 15, 2024. Retrieved February 21, 2024.
  33. ^ McCracken, Harry (February 15, 2024). "Google's new Gemini 1.5 AI can dive deep into oceans of video and audio". Fast Company. Archived from the original on February 17, 2024. Retrieved February 21, 2024.
  34. ^ Stokes, Samantha (February 15, 2024). "Here's everything you need to know about Gemini 1.5, Google's newly updated AI model that hopes to challenge OpenAI". Business Insider. Archived from the original on February 19, 2024. Retrieved February 21, 2024.
  35. ^ Khan, Jeremy (February 21, 2024). "Google unveils new family of open-source AI models called Gemma to take on Meta and others—deciding open-source AI ain't so bad after all". Fast Company. Archived from the original on February 21, 2024. Retrieved February 21, 2024.
  36. ^ Alba, Davey (February 21, 2024). "Google Delves Deeper Into Open Source with Launch of Gemma AI Model". Bloomberg News. Archived from the original on February 21, 2024. Retrieved February 21, 2024.
  37. ^ Metz, Cade; Grant, Nico (February 21, 2024). "Google Is Giving Away Some of the A.I. That Powers Chatbots". The New York Times. ISSN 0362-4331. Archived from the original on February 21, 2024. Retrieved February 21, 2024.
  38. ^ Elias, Jennifer (August 12, 2024). "Google rolls out its most powerful AI models as competition from OpenAI heats up". CNBC. Archived from the original on May 14, 2024. Retrieved August 13, 2024.
  39. ^ "Gemma 2 is now available to researchers and developers". Google. June 27, 2024. Retrieved August 15, 2024.
  40. ^ "Updated production-ready Gemini models, reduced 1.5 Pro pricing, increased rate limits, and more- Google Developers Blog". developers.googleblog.com.
  41. ^ "Introducing Gemini 2.0: our new AI model for the agentic era". Google. December 11, 2024. Retrieved December 20, 2024.
  42. ^ "The next chapter of the Gemini era for developers- Google Developers Blog". developers.googleblog.com. Retrieved December 20, 2024.
  43. ^ "Gemini 2.0 Flash (experimental) | Gemini API". Google AI for Developers. Retrieved December 20, 2024.
  44. ^ Gemini: A Family of Highly Capable Multimodal Models (PDF) (Technical report). Google DeepMind. December 6, 2023. Archived (PDF) from the original on December 6, 2023. Retrieved December 7, 2023.
  45. ^ Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context (PDF) (Technical report). Google DeepMind. February 15, 2024. Archived (PDF) from the original on February 26, 2024. Retrieved May 17, 2024.
  46. ^ Gemma Team; Riviere, Morgane; Pathak, Shreya; Sessa, Pier Giuseppe; Hardin, Cassidy; Bhupatiraju, Surya; Hussenot, Léonard; Mesnard, Thomas; Shahriari, Bobak (August 2, 2024), Gemma 2: Improving Open Language Models at a Practical Size, arXiv:2408.00118
  47. ^ "Gemma explained: An overview of Gemma model family architectures- Google Developers Blog". developers.googleblog.com. Retrieved August 15, 2024.
  48. ^ "PaLI: Scaling Language-Image Learning in 100+ Languages". research.google. Retrieved August 15, 2024.
  49. ^ a b Heikkilä, Melissa; Heaven, Will Douglas (December 6, 2023). "Google DeepMind's new Gemini model looks amazing—but could signal peak AI hype". MIT Technology Review. Archived from the original on December 6, 2023. Retrieved December 6, 2023.
  50. ^ howdhury, Hasan (August 29, 2023). "AI bros are at war over declarations that Google's upcoming Gemini AI model smashes OpenAI's GPT-4". Business Insider. Archived from the original on August 29, 2023. Retrieved September 7, 2023.
  51. ^ Harrison, Maggie (August 31, 2023). "OpenAI Rages at Report that Google's New AI Crushes GPT-4". Fortune. Archived from the original on August 31, 2023. Retrieved September 7, 2023.
  52. ^ Musk, Elon [@elonmusk] (August 29, 2023). "Are the numbers wrong?" (Tweet). Archived from the original on September 1, 2023. Retrieved October 15, 2023 – via Twitter.
  53. ^ Langley, Hugh (October 12, 2023). "Google VP teases Gemini's multimodal future: 'I've seen some pretty amazing things.'". Business Insider. Archived from the original on October 12, 2023. Retrieved October 15, 2023.
  54. ^ Madden, Michael G. (December 15, 2023). "Google's Gemini: is the new AI model really better than ChatGPT?". The Conversation. Archived from the original on December 15, 2023. Retrieved February 4, 2024.
  55. ^ Sullivan, Mark (December 6, 2023). "Gemini-powered Google phones may make Siri even more of an Achilles' heel for the iPhone". Fast Company. Archived from the original on December 7, 2023. Retrieved December 7, 2023.
  56. ^ Soni, Aditya (December 7, 2023). "Alphabet soars as Wall Street cheers arrival of AI model Gemini". Reuters. Archived from the original on December 7, 2023. Retrieved February 4, 2024.
  57. ^ Swartz, Jon (December 7, 2023). "Gemini, Google's long-awaited answer to ChatGPT, is an overnight hit". MarketWatch. Archived from the original on December 7, 2023. Retrieved February 4, 2024.
  58. ^ Elias, Steve Kovach, Jennifer (December 8, 2023). "Google faces controversy over edited Gemini AI demo video". CNBC. Archived from the original on December 9, 2023. Retrieved December 9, 2023.{{cite web}}: CS1 maint: multiple names: authors list (link)

Further reading

[edit]
[edit]
Retrieved from "https://wiki.riteme.site/w/index.php?title=Gemini_(language_model)&oldid=1265056374"