Key Takeaways
- Google is using an AI previously developed to beat a Go champion to improve its natural language AI model
- The company’s biggest AI teams are coming together to bring AI to the next level
- Google AI scientists believe they can beat OpenAI’s ChatGPT
Since ChatGPT’s release on November 30th last year, the AI language model has taken over every headline. The program’s name has become synonymous with any mention of Artificial Intelligence as it continued to gain popularity.
The program created by OpenAI, now part of Microsoft, continues to beat records, including gathering 100 million users within 3 months. That number of users was gained by Tiktok in 9 months and by Instagram in over two and a half years.
Most interestingly, due to the nature of the program’s natural language model, enthusiasts even started wondering if ChatGPT could overtake Google one day in the search industry. Google has been dominating internet search for over 20 years.
ChatGPT even forced Google to reveal its own language model (seemingly in a rushed manner) called Bard. As users tested Bard, the language model seemed inferior to its main competition. Although Bard offered features that ChatGPT does not possess, it failed to takeover has the household name AI language model.
Google now is switching strategies, bringing its biggest AI teams together to work on a project codenamed Gemini. Gemini will use an AI model previously used to beat a boardgame world champion to build upon the company’s already-advanced AI tools, in hopes of taking the crown off ChatGPT’s head.
Gemini Is A Go
AlphaGo , an AI model used to defeat Go world champions Lee Sedol and Ke Jie, was created by DeepMind, an AI company created by Demis Hassabis. Deepmind was acquired by Google in 2014 to help the company improve its native tools, such as Gmail, Photos, and Google Docs Suite.
Google already had its own AI lab, Brain. The two teams are now working together on a project the company is codenaming Gemini .
“Gemini was created from the ground up to be multimodal, highly efficient at tool and API integrations, and built to enable future innovations, like memory and planning. While still early, we’re already seeing impressive multimodal capabilities not seen in prior models,” – said Sundar Pichai, Google CEO at Google’s 2023 I/O conference.
During the conference, Pichai displayed a wide array of tools developed with the help of artificial intelligence, including improved email composition on Gmail, improved navigation through Maps, as well as a mind-boggling tool called Magic Editor.
During the presentation, Pichai showed how Magic Editor can manipulate any photograph taken and saved on the company’s Photos app using a few taps. When the tool gets released to the public, users will be able to edit their photos by changing the look of the sky or bringing the main subject to the center of the frame, without jeopardizing the quality of the image.
PaLM 2
Alongside Bard, Magic Editor, Gemini, and all of the improvements coming to Google tools, the company has been working on its new Pathways Language Model (PaLM 2). PaLM 2 is the company’s most recent Large Language Model (LLM) that is more capable of advanced reasoning, coding, and mathematics than its predecessor.
“PaLM 2 models deliver excellent foundational capabilities across a wide range of sizes. We’ve affectionately named them Gecko, Otter, Bison, and Unicorn. Gecko is so lightweight that it can work on mobile devices: fast enough for great interactive applications on-device, even when offline.
PaLM 2 models are stronger in logic and reasoning thanks to broad training on scientific and mathematical topics. It’s also trained on multilingual text — spanning more than 100 languages — so it understands and generates nuanced results.”
Google is pushing the boundaries of what these AI tools are capable of by enabling PaLM 2 to generate medical information using medical imaging like plain films and mammograms.
The idea is that PaLM 2 can be a useful companion for medical professionals, helping them process medical data quicker and provide more accurate care for patients.
Google Vs ChatGPT
The problem was company announcements regarding new technologies is the lack of real-world application experience. ChatGPT has already proven itself useful and popular among millions of users around the world while improving its own model through GPT-4. Google’s Bard however has yet to gather the same fandom ChatGPT has been enjoying.
OpenAI CEO Sam Altman said that both ChatGPT and Dall-E, the company’s image-generating AI tool, took years and over $100 million to develop.
However, Altman does admit that models in the future will not require as much time, money, or effort to develop their transformers. Transformers are the databases built for AI tools to develop their knowledge bases, enabling them to answer user queries.
Google is still the world’s #1 software company. And Google Search is still the world’s go-to search module. But, the company will have to work on developing its tools, such as Bard and Gemini in more efficient and effective ways if it wishes to takeover the AI commercial tools market.
