Google Gemini 1.5 Pro is a new, more efficient model of artificial intelligence

Thursday Google unveil Gemini 1.5 Pro, which the company describes as offering “significantly enhanced performance” compared to the previous model. The company's AI path — seen internally as critical to its future — follows last week's unveiling of the Gemini 1.0 Ultra, along with the rebranding of the Bard chatbot (to Gemini) to align with the new model's more powerful and versatile capabilities.

In an announcement blog post, Google CEO Sundar Pichai and Google DeepMind CEO Demis Hassabis attempt to balance reassuring their audiences about the ethical safety of AI while touting the capabilities of their rapidly evolving models. “Our teams continue to push the boundaries of our latest models with safety at their core,” Pichai summed up.

The company needs to stress the safety of AI skeptics (including a former Google executive) and government regulators. But it also needs to emphasize the accelerating performance of its models to AI developers, potential customers, and investors who worry that the company has been too slow to respond to OpenAI's success with ChatGPT.

Pichai and Hassabis say the Gemini 1.5 Pro delivers similar results to the Gemini 1.0 Ultra. However, Gemini 1.5 operates at this level more efficiently, with lower computational requirements. Multimedia capabilities include processing text, images, videos, audio, or code. As AI models advance, they will continue to offer a more diverse set of capabilities in a single claim box (another recent example is the integration of OpenAI to generate DALL-E 3 images into ChatGPT).

Gemini 1.5 Pro can also handle up to 1 million tokens, or its AI-powered data modeling modules can process them in a single request. Google says the Gemini 1.5 Pro can process more than 700,000 words, an hour of video, 11 hours of audio, and codebases with more than 30,000 lines of code. The company says it has “successfully tested” a version that supports up to 10 million tokens.

The company says Gemini 1.5 Pro maintains high accuracy on queries with larger numbers of tokens when it has more new data to learn. She says she liked the model Needle in a haystack evaluation. In this test, developers insert a small piece of information within a long block of text to see if the AI ​​model can capture it. Google said Gemini 1.5 Pro can find embedded text 99 percent of the time in data blocks up to 1 million tokens long.

Google says the Gemini 1.5 Pro can reflect on various details from the 402-page Apollo 11 moon mission transcripts. Additionally, it can analyze plot points and events from an uploaded 44-minute silent movie starring Buster Keaton. “Because the 1.5 Pro's long context window is the first of its kind among large-scale models, we are constantly developing new evaluations and benchmarks to test its new capabilities,” Hassabis wrote.

Google launches Gemini 1.5 Pro with 128,000 token capabilities same Number That's where OpenAI's (publicly announced) GPT-4 models reach their limit. Hassabis says Google will eventually introduce new pricing tiers that support up to 1 million unique queries.

Gemini 1.5 Pro is also adept at learning new skills from information in long prompts – without additional tuning (“learning in context”). In a standard called Machine translation of one book,The model learned a grammar guide for Kalamang, a language spoken by fewer than 200 speakers globally and had never been trained on before. The company says the Gemini 1.5 Pro learns performance at a similar level as a human learns the same content when translating English to Kalamang.

In a part of the announcement that will catch developers' attention, Google says the Gemini 1.5 Pro can perform problem-solving tasks across longer code blocks. “When given a prompt containing more than 100,000 lines of code, they can better reason through examples, suggest useful modifications and provide explanations for how different pieces of code work,” Hassabis wrote.

On the ethics and safety front, Google says it's taking “the same approach to responsible publishing” as it did with the Gemini 1.0 models. This includes developing and applying red teaming techniques, where a group of ethical developers essentially act as devil's advocates, testing “a range of potential harms.” Additionally, the company says it performs heavy scrutiny in areas such as content integrity and representational damage. The company says it continues to develop new ethical and safety tests for its AI tools.

Google releases Gemini 1.5 early access for developers and enterprise customers. The company plans to eventually make it more widely available. Gemini 1.0 is currently available to consumers, along with a Professional alternative Which costs $20 per month.