Google has released a new Pro version of its latest AI, Gemini, and according to company sources, it outperformed GPT-3.5 (the free version of ChatGPT) in extensive testing.
Gemini Ultra’s Impressive Performance on Academic Benchmarks
Gemini Ultra outperforms current state-of-the-art results on 30 of the 32 widely-used academic benchmarks used in large language model (LLM) research and development, according to performance reports.
Gemini’s Multimodal Capabilities
Google has been accused of trailing OpenAI’s ChatGPT, which is widely regarded as the most popular and powerful AI platform. According to Google, Gemini was trained to be multimodal, which means it can process various types of media such as text, images, video, and audio.
Gemini’s Breakthrough in Massive Multitask Language Understanding
Insider also reports that, with a score of 90.0%, Gemini Ultra is the first model to outperform human experts on MMLU (massive multitask language understanding), which tests both world knowledge and problem-solving abilities by combining 57 subjects such as math, physics, history, law, medicine, and ethics.
Gemini’s Availability and Access
The Google-based AI is available in three sizes, or stages, for the Gemini platform: Ultra, Pro, and Nano (designed for mobile devices). According to TechCrunch, the company will make Gemini Pro available to enterprise customers via its Vertex AI program, as well as developers via AI Studio, on December 13. According to reports, the Pro version is also accessible through Bard, the company’s chatbot interface.
Gemini’s Ability to Comprehend Nuanced Information
According to Eli Collins, VP of product at DeepMind (the division of Google tasked with building out the AI platform), Gemini Ultra can comprehend “nuanced” information in text, images, audio, and code.
Concerns About Data Sources and Consent
Collins revealed that some of the data used to create the app came from public web sources. However, the company did not address the sources for training the AI directly.
The Copyright and Fair Use Challenge for AI Platforms
As more capabilities emerge, “consent” is becoming more important for Gemini and other AI offerings. Aside from performance achievements, what about the raw materials required to make these platforms powerful? I’m not referring to energy, hardware, or chips. ChatGPT and other AI platforms are trained (and use) the works of millions of artists, inventors, educators, and authors. Content creators are suing Microsoft, GitHub, OpenAI, and Stability AI, alleging unfair usage. According to the Associated Press, famous authors such as John Grisham (The Firm) and George R.R. Martin (Game of Thrones) are among 17 writers who are suing Open AI for “systemic theft on a mass scale.”
According to the US Copyright Office, it is permissible to use limited portions of a work, including quotes, for purposes such as commentary, criticism, news reporting, and scholarly reports under the fair use doctrine of the US copyright statute. This article does, in fact, use pull quotes with attribution to sources. The problem for Gemini, ChatGPT, and other AI solutions is that the original human creators have not “granted permission.” There is no remuneration for using their entire body of work to create something new. These authors and creators serve as de facto AI trainers, with no compensation for their time and efforts.
Frequently Asked Questions (FAQs)
1. What is Gemini, and how does it compare to ChatGPT?
Gemini is Google’s latest AI model, and according to company sources, it outperformed ChatGPT (GPT-3.5) in extensive testing. It excelled in 30 out of 32 widely used academic benchmarks.
2. What sets Gemini apart from other AI models?
One of Gemini’s distinguishing features is its multimodal capability. It can process various types of media, including text, images, video, and audio, making it a versatile AI model.
3. What is MMLU, and why is Gemini Ultra’s performance significant in this area?
MMLU stands for Massive Multitask Language Understanding, a test that combines 57 subjects, including math, physics, history, law, medicine, and ethics. Gemini Ultra achieved a score of 90.0% on this test, surpassing human experts, which is a significant milestone.
4. When and how will Gemini be available to users?
Gemini is available in three sizes: Ultra, Pro, and Nano (for mobile devices). Google plans to make Gemini Pro accessible to enterprise customers through its Vertex AI program and developers via AI Studio. It’s expected to be available on December 13.
5. How does Gemini comprehend nuanced information, and what can it process?
Eli Collins, VP of product at DeepMind, the division of Google behind Gemini, stated that Gemini Ultra can comprehend nuanced information in text, images, audio, and code.
6. What are the concerns about data sources and consent raised in the article?
The article raises concerns about the sources of data used to train AI models like Gemini and the lack of explicit consent from content creators whose work contributes to the training data. It discusses the challenges related to using copyrighted material without permission and compensation.
7. What is the fair use doctrine, and how does it relate to AI models like Gemini and ChatGPT?
The fair use doctrine, as mentioned in the article, allows limited use of copyrighted works, such as quotes, for purposes like commentary, criticism, news reporting, and scholarly reports. However, the challenge for AI models like Gemini and ChatGPT is that they use entire bodies of work without explicit permission or compensation to content creators.
8. Are there legal actions mentioned in the article related to the use of copyrighted material in AI models?
Yes, the article mentions legal actions taken by content creators, including famous authors like John Grisham and George R.R. Martin, who are suing OpenAI for alleged “systemic theft on a mass scale” related to the use of their copyrighted material in AI training.
9. What is the article’s stance on the ethical and legal aspects of using copyrighted material in AI development?
The article highlights the growing importance of obtaining consent and fairly compensating content creators whose works are used to train AI models, raising ethical and legal questions about the practice.
10. How can individuals or organizations stay informed about developments in AI and related legal issues?
To stay informed about AI developments and legal issues, it is advisable to follow news sources, legal updates, and official statements from AI developers and organizations involved in AI research and development.
