Happy Wednesday!

In Today’s Email:

• Meet Med-PaLM 🩻

• Google Assistant is getting an overhaul 🦮

• LLM 🤝 Massive APIs

Meet Med-PaLM 🩻

LLMs will forever influence how clinicians interact with data.

Researchers at Google have continued to build towards their ultimate goal, creating a general-purpose biomedical AI system. To build this system, researchers from Google Research and Google DeepMind have collaborated to create MultiMedBench, a benchmarking system comprised of 14 different biomedical activities to guide the development of the biomedical AI system.

Authors from Google have recently released a proof of concept called Med-PaLM Multimodal. The model can decipher biomedical data including, medical pictures, clinical language, and genetic code. Med-PaLM Multimodal performed exceptionally on the MultiMedBench assessment, in some cases, it even outperformed specialized models.

The team working on Med-PaLM Multimodal has summarized their contributions:

1. The work demonstrates the potential of generalist biomedical AI systems for medical applications, though access to extensive biological data for training and validating in-use performance continues to be a problem.

2. MultiMedBench has 14 unique tasks encompassing a range of biomedical modalities. Med-PaLM M, the first multitasking generalist biomedical AI system, has been introduced that does not require task-specific modifications.

3. The AI system demonstrates emergent abilities, such as generalization to new medical concepts and zero-shot medical reasoning.

4. Potential clinical utility is indicated by a human review of Med-PaLM M’s outputs, particularly in producing chest X-ray reports.

5. With low average mistakes, radiologists favor Med-PaLM M reports over radiologists’ reports in up to 40.50% of cases.

Check out the paper here.

Google Assistant is getting an overhaul 🦮

Generative AI will drastically change the way Google Assistant functions.

According to an internal email sent to employees on Monday, Google plans to redesign its Assistant to focus on generative AI. The transition will alter how Assistant works for developers, consumers, and employees within the company. Google says some of the work on the new Assistant has already begun.

Google plans on reorganizing the teams that work on Assistant, which will include a small number of layoffs in conjunction with the transition. According to the email, dozens of jobs will be eliminated out of the thousands of employees working on the Assistant.

“We remain deeply committed to Assistant and we are optimistic about its bright future ahead,” Peeyush Ranjan, Google VP and director of the product says. This comes as more companies such as Amazon begin to scale back their traditional digital assistants in favor of AI-powered tools.

LLM 🤝 Massive APIs

The first LLM specializing in writing API calls is here.

Gorilla is a LLaMA-based Large Language Model designed by researchers from UC Berkeley and Microsoft Research. The model is designed to write API calls based on text prompts, it can already invoke 1,600+ API calls while reducing hallucination. Although LLMs excel at a variety of tasks including program synthesis and mathematical reasoning, their ability to write APIs has been lacking.

The model is trained on three massive machine-learning datasets: Hugging Face, Tensor Flow Hub, and Torch Hub. The researchers behind the model are also planning to add new domains such as Kubernetes, GCP, AWS, OpenAI, and more.

Gorilla has demonstrated its ability to surpass the performance of GPT-4 on writing API calls. The model also has the ability to enable flexible API updates and version changes rapidly when combined with a document retriever. It already has 6k stars on GitHub, which you can check out here.

That’s it for today.

We’re switching to three days a week, MWF, see you Friday!