The arrival of Google's AI model Gemini 1.0 has heralded a new phase in the AI ​​race between Microsoft/OpenAI, Meta and Google. Elon Musk is also joining the fray.

 

 

While the previous version of Google's AI chatbot Bard scored relatively poorly, Bard's successor seems promising. Although previously postponed until early 2024, Google has now accelerated its release with a new version of its own GenAI under the name Gemini. It looks like Gemini is even overtaking the latest version of ChatGPT, because of how striking it is multimodal performance.

Multimodal: opportunities for customer service

Multimodal and generative AI in the case of Google Gemini means that the AI ​​solution can watch and listen to the user, but also produce all kinds of audio itself. The user can also ask Gemini questions about still and moving images, where Gemini also reasons logically. This is how Gemini is able to play the game “ball ball" to solve.

This logical reasoning becomes interesting for customer contact if a link can be made with a knowledge base that also contains videos and photos; For example, the chatbot could then independently help a customer solve technical issues such as connecting a router.

Also read:

Google demo video about Gemini turns out to be partly fake

 

 

Nano, Pro and Ultra

Gemini is based on the PaLM 2 model, Google's proprietary LLM. Gemini is released in three different versions, of which Nano (the lightest variant) can run on smartphones, Pro is presented as a scalable version and Ultra as the top variant. Of these, Google itself says that this model is the one that “outperforms human experts in MMLU (massive multitask Language Understanding), which combines 57 subjects, including math, physics, history, law, medicine and ethics, for testing both general knowledge and problem-solving skills.” This last variant still needs further testing and is presented by Google as the enterprise solution.

Phased rollout

From now on, chatbot Bard will also run on Gemini Pro, which allows it to reason, plan and understand at a higher level, but for the time being only in English. Gemini will also be integrated into other Google services such as search and added to Google's Pixel smartphones. This immediately makes functions available such as Summarizing in the Recorder app and Smart Reply for WhatsApp. With these types of functionalities, Gemini comes close to Microsoft's Copilot.

Gemini 1.0 will be available to developers on December 13. to custom AI products via the Gemini API in Google AI Studio or Google Cloud Vertex AI.

AI race

Meanwhile, the AI ​​race is expanding. In addition to Google's LLMs, Meta (LLaMA) and OpenAI (ChatGPT), Elon Musk is also joining the fray. He has announced a funding round with a target size of $1 billion. That money is needed for xAI, a start-up with which Musk wants to compete with the other players. Amazon also recently launched its own platform: Amazon Q.

General purpose technology

The fact that generative AI is getting better is also leading to renewed discussions about the impact on employment. According to Utrecht professor and economist Anna Salomons, the arrival of language models such as ChatGPT and Bard heralds a new era. They make AI one general purpose technology, just like the steam engine: it can be used for various purposes with minimal adjustments. Salomons expects that AI will mainly have a leveling effect: a positive impact on lower-educated people and a negative impact on (the work of) higher-educated people. (Ziptone/Google)

Featured, Technology