Last week, Google hosted its annual I/O developer conference, showcasing the company’s vision for its products and services. This usually entails exciting conversations about Google’s mobile platform, Android, or the latest devices that the company plans to release. However, this year, something else stole the show: Google’s amazing advances in artificial intelligence (AI).
Artificial intelligence has been a sensation both in Silicon Valley and around the world over the past few months, especially after the release of Chat-GPT and advanced generative AI systems. This has somewhat led to a frantic race among the tech giants: everyone wants a piece of the AI pie, and the race to the top is intense.
For example, Microsoft is investing significant resources in its AI capabilities, relentlessly trying to integrate the technology across a variety of sectors, from manufacturing to healthcare. Amazon is also innovating in this area, offering customers turnkey services to build their own machine learning and AI models.
In a similar fashion, Google has made amazing strides in its AI journey. Although many may suspect that the hype around AI is relatively new, the company has already been laying the groundwork for advanced language models and usable AI for many years. This work culminated in products with amazing potential, including PaLM 2, Google’s latest generation of Large Language Models.
As an introduction to PaLM 2, Zoubin Ghahramani, Vice President of Google DeepMind, wrote last week: “Today we introduce PaLM 2, the next-generation language model. PaLM 2 is a state-of-the-art language model with enhanced multilingual, logic, and coding capabilities.” He goes on to vividly describe these capabilities. Bigger, and discusses how PaLM 2 is trained across 100 languages, has higher ability in advanced logical reasoning, math, and logic, and can even create advanced code.
Notably, Ghahramani also discusses one of the most anticipated areas of impact for AI: healthcare. He delves into Med-PaLM 2, the large language model specifically developed to generate medical insights: “Trained by our health research teams with medical knowledge, [Med-PaLM 2] It can answer questions and summarize insights from a variety of thick medical texts. It achieved the latest results in Medical Proficiency, and was the first major language model to operate at an “expert” level similar to the US Medical Licensing Examination [USMLE] Questions. We are now adding multimodal capabilities to aggregate information such as x-rays and mammograms to one day improve patient outcomes. Med-PaLM 2 will open up to a small group of cloud customers for feedback later this summer to identify safe and beneficial use cases.”
The full 2023 keynote can be viewed here:
Google Cloud, especially for healthcare, has been incredibly successful over the past few years. For example, the Google Cloud Healthcare Data Engine is an industry-leading platform for improving interoperability in healthcare, which can solve some of the sector’s most challenging problems: massive data fragmentation, lack of coherent insights, and fragmented patient journeys.
“Industry-tailored LLMs such as Med-PaLM 2 are part of a thriving family of generative AI technologies that have the potential to enhance human experiences,” explains Google Cloud Global Director of Healthcare Strategy and Solutions, Aashima Gupta, and Global Director of Health Plan Strategy and Solutions, Amy Waldron. healthcare in a big way.We look forward to working with our clients to understand how Med-PaLM 2 can be used to facilitate informative discussions, answer complex medical questions, and find insights in complex and unstructured medical texts.They may also explore its usefulness to help craft short and long-form responses and summarize documents and ideas from internal datasets and scientific knowledge collections.”
The Cloud team was also responsible for launching the Medical Imaging Suite, a comprehensive AI platform for making medical imaging data “available, interoperable, and useful.”
Undoubtedly, the development of PaLM 2 will be a game-changer in healthcare. Whether it is used to perform advanced analytics of massive amounts of public health or patient data, or simply as a way to triage and improve a physician’s workflow, technology has huge potential to make a measurable impact.
This technology could be very useful to organizations that want to scale their tactical approach to improving patient outcomes, particularly in terms of synthesis and insight generation. For example, for public entities, harnessing this technology may one day make it possible to ingest terabytes of otherwise unstructured public health data to generate usable insights. For private organizations, this technology could prove to be a huge boon in the areas of interoperability, improving patient journeys, and success in true longitudinal care.
However, this technology requires guardrails. While there is great potential for beneficial use, there is also some “fear of the unknown”. Meaning that in the wrong hands, the powerful technology driving these advances in AI could have negative repercussions. This is the main reason why the developer conference had an entire segment dedicated to the “responsible development” of AI, as a reminder of the checks and balances required in this rapidly growing field of technology. However, if developers and innovators are able to create and harness this technology in a safe and sustainable way, it could change the face of healthcare delivery for generations to come.