Google actually beat GPT-4 this time? Gemini Ultra released

Fireship
8 Feb 202405:04

Summary

TLDRIn this video script, the host discusses Google's new AI model, Gemini Advance, which has been renamed from Bard and claims to rival GPT 4. The script covers the model's speed, response quality, safety features, and political bias, as well as its capabilities in code reading, writing, and execution. It also touches on the extensions available for both Gemini and GPT 4, concluding that Gemini is on par with, if not slightly better than, GPT 4, setting the stage for an AI rivalry.

Takeaways

  • πŸ“’ Google has renamed Bard to Gemini and introduced Gemini Advance, featuring the Gemini Ultra model, as a response to the controversy around its AI capabilities.
  • πŸ’Έ Access to the advanced Gemini Ultra model requires a subscription fee, raising ethical questions about supporting surveillance capitalism versus open AI initiatives.
  • ⏱️ Gemini is significantly faster than its predecessor, offering at least two to three times the speed in responses.
  • πŸ“ In a subjective test, Gemini produced a poem about JavaScript in the style of Bukowski, which was considered superior to other AIs tested.
  • πŸ›‘οΈ Google emphasizes that Gemini has strong safety features and is 'woke', refusing to engage in harmful or politically biased activities.
  • 🚫 Gemini showed a reluctance to generate content that could be seen as promoting violence or negativity, such as images of computer destruction.
  • πŸ€– There is an indication of political bias in Gemini's responses, as it provided a paragraph for one political figure but not for another, citing the inability to write opinions as facts.
  • πŸ–ΌοΈ When it comes to image generation, Gemini's results were mediocre, suggesting that other tools like mid Journey or stable diffusion XL are better for high-quality AI images.
  • πŸ’» Gemini and GPT 4 demonstrated similar capabilities in understanding and explaining code, but Gemini lacked the 'package main' in its code output, requiring manual correction.
  • πŸ”— Gemini's code output includes links to relevant code, adding transparency about the data used for training the model.
  • πŸ›οΈ Gemini has extensions that are not yet open to developers, unlike GPT 4's agent marketplace, which allows for plugin extensions by developers.

Q & A

  • What claim did Google make a few months ago regarding its AI?

    -Google claimed that it had built an AI superior to GPT 4.

  • What was the controversy surrounding the video Google released featuring Gemini Ultra?

    -The video was controversial because it was revealed to be mostly fake, showing a conversation with Gemini Ultra that appeared human-like.

  • What is the name of the new AI model released by Google, and what does it aim to reflect?

    -The new AI model is called Gemini Advance, which reflects its underlying model and is designed to be the most advanced large language model in the world.

  • What is the monthly cost for accessing the Gemini Ultra model through a Google One plan?

    -The cost for accessing Gemini Ultra is $20 a month through a Google One plan, which also includes 2 terabytes of drive storage and other Google workplace features.

  • How does the speed of Gemini Advance compare to GPT 4?

    -Gemini Advance is significantly faster, being at least two or three times faster than GPT 4.

  • What subjective test was used to compare the response quality of Gemini Advance and GPT 4?

    -The subjective test involved asking the AI to write a poem about JavaScript in the style of Charles Bukowski.

  • What was the main reason Google did not release Gemini Ultra a few months ago?

    -Google did not release Gemini Ultra due to safety concerns, but now claims it is the safest and 'wokest' AI available.

  • How did Gemini Advance respond when tested for political bias?

    -Gemini Advance showed a potential political bias when asked about Barack Obama and Donald Trump, providing a paragraph for one but refusing to write opinions as fact for the other.

  • What is the context length of Gemini Advance and GPT 4 Turbo, and what does this theoretically indicate about their capabilities?

    -Gemini Advance has a context length of 32,000 tokens, while GPT 4 Turbo has 128k, theoretically indicating that GPT 4 should be able to write better code in the context of a massive code base.

  • What is the main advantage of GPT 4's new agent marketplace over Gemini Advance's extensions?

    -GPT 4's agent marketplace allows developers to extend its capabilities with their own plugins, while Gemini Advance's extensions are currently not open to developers and are all Google-based.

  • What is the final verdict of the video on whether Gemini Advance can kill Chat GPT?

    -The video concludes that Gemini Advance is on par with, if not slightly better than, Chat GPT, but it is not revolutionary and the competition between the two is expected to continue.

Outlines

plate

This section is available to paid users only. Please upgrade to access this part.

Upgrade Now

Mindmap

plate

This section is available to paid users only. Please upgrade to access this part.

Upgrade Now

Keywords

plate

This section is available to paid users only. Please upgrade to access this part.

Upgrade Now

Highlights

plate

This section is available to paid users only. Please upgrade to access this part.

Upgrade Now

Transcripts

plate

This section is available to paid users only. Please upgrade to access this part.

Upgrade Now
Rate This
β˜…
β˜…
β˜…
β˜…
β˜…

5.0 / 5 (0 votes)

Related Tags
AI ModelsGemini UltraGPT 4AI EthicsTechnical TestCode GenerationAI ComparisonSurveillance CapitalismProductivity ToolsAI SafetyProgramming AI