Google is set to launch major new advancements to its AI portfolio with its upcoming I/O conference, including a large language model named PaLM 2, improvements to its chatbot Bard, and a new powerful AI core model. PaLM 2 is trained in more than one hundred languages and is considered to be a major leap in AI understanding compared to current popular models such as OpenAI’s GPT-3.
It is expected that Google will demonstrate significant updates to its chatbot Bard, powered by PaLM 2. The updates will include a larger version known as ‘Multi-Bard’ with logically-sound reasoning capabilities and advanced math. An impressive feature of these new updates will be the introduction of multimodal AI models. This will allow for complex user interactions similar to OpenAI’s GPT-4, which can process images, analysing sound and videos, as well as natural language.
Sundar Pichai, CEO of Alphabet, announced in April this year that Google would develop its own multimodal AI models to compete with OpenAI’s GPT-4, whose benchmark outperformed PaLM 2 in the Multi-task Language Understanding (MMLU) test.
Google DeepMind’s CEO, Demis Hassabis, has also been given the responsibility to develop the “most capable and responsible AI systems” in the introduction of Google’s new unified AI division.
The firm has said that improving the chatbot’s performance is a priority and seek to reduce the “hallucinations” which have plagued popular chatbot models. This means producing more factual answers, as was the case for GPT-4. Med-PaLM, for example, achieved a score of 85% at the MedQA dataset of the Medical Licensing Examination.
Users of Bard will also be able to expect gains in its ability to sustain conversations over multiple sessions, as well as improved code generation. Google Lens will also be integrated with the latest model, allowing users to flowcharts to be converted into subtle code. This will greatly benefit tech corporations and developers utilizing GitHub Copilot.
Google’s algorithm and AI advancements are being widely discussed and it will be fascinating to see how these new updates will impact the world of artificial intelligence. The drive to utilize the most proficient and responsible AI systems will undoubtedly be a key focus in the company’s journey to improve its AI portfolio.
