What do you see in this picture?

It was a day of Walmart Health's demise, but Walmart itself didn't suffer much, considering it was a down day for the market on the heels of the 'hot' labor data.

What I see in this picture is that markets heard loud and clear (among other

Clinical Workflow Critic Agreed. Another issue is the transparency and interpretability of AI models in healthcare. Every clinician keeps saying, 'I'm not going to use this AI until I know exactly why it came up with this answer,' and I appreciate the sentiment. However, I don't think many people realize

Yesterday, Google released its brand new Gemini model for medicine, aptly named #MedGemini .

In my rebuttal to the paper, as part of the AI Health Uncut newsletter (link in profile), I explain why adding more GPUs to your AI models may not be what clinicians and patients need.

Delighted to share ✨Med-Gemini✨ - our new family of multimodal models for medicine unlocking new possibilities for health - arxiv.org/pdf/2404.18416

More accurate multimodal conversations about medical images🩻, surgical videos📽️, genomics🧬, ultra-long health records📚, ECGs🫀

Google lays off its entire #Python team.

My thoughts:

1️⃣ Google's decision to lay off its talented workers may backfire, as their competitors will be happy to hire them.

2️⃣ #Google may have just been 'McKinseyed' - that is, handed off the most important decisions about the

I provide my reasons for the Optum telehealth exit and explain the state of telehealth in my latest newsletter (link in my profile):

1. Just like with Teladoc Health and Amwell, overspending and overinvestment during the heydays of COVID (in this case, 2021), backfired.

2.

Christina Farr I provide my reasons for the Optum telehealth exit and explain the state of telehealth in my latest newsletter (link in my profile):

1. Just like with Teladoc and Amwell, overspending and overinvestment during the heydays of COVID (in this case, 2021), backfired.
2. When you own

Real Doc Speaks Using foundational AI models (such ChatGPT, Gemini, and even Google's Med PaLM 2) for medicine is irresponsible and dangerous. These models were not built (not to mention, tested) for healthcare, especially for precision medicine and diagnostics. Corporations will continue

Real Doc Speaks Epic is next. It's not a question of if, but when. Those who have tried to work with Epic's API endpoints— 'tried' because Epic creates all sorts of hurdles to prevent third parties from accessing their system, even when providers, i.e., clients, demand that access—know that Epic

Blake Madden 🏥 From what I’m seeing, physicians are increasingly adopting telehealth using tools that are already installed on their computers, such as their EHR systems or Doximity.

Lighthouse Direct Primary Care If hackers are infiltrating Change, the next obvious target is EHR systems, which should be a walk in the park for cybercriminals. However, in a weird way, it might not be as easy because many of those systems are very low-tech. Hacking a fax machine is much harder than hacking a

Lighthouse Direct Primary Care There is pretty much certainty that the medical data of all Americans has been hacked. Cyber attacks on American healthcare occur daily. In 2023 alone, 480 healthcare data breaches were reported, impacting 87 million patients. 2024 is on track to surpass these figures,

Dr. Sam Madeira, NMD I agree that AI should be proliferated in hospitals with close collaboration with nurses. However, looking at the photos, some might conclude that nurses are against AI altogether. If nurses are concerned about patient outcomes, AI should absolutely be part of the solution.

Today at Answer.AI we've got something new for you: FSDP/QDoRA. We've tested it with AI at Meta Llama3 and the results blow away anything we've seen before.

I believe that this combination is likely to create better task-specific models than anything else at any cost. 🧵

Real Doc Speaks On the contrary, we need to thank hospitals and EHR systems for ignoring the criticism and blocking all the tech hysteria of the last three decades. They steadily stayed the course, refusing to update the DOS operating system of the 1990s. By staying very low tech for so long,

Here is a paper that finally answers a question: Which method is better for RLHF/supervised fine-tuning, DPO or PPO?

#Llama3 shatters stereotypes!

Introducing Meta Llama 3: the most capable openly available LLM to date.

Today we’re releasing 8B & 70B models that deliver on new capabilities such as improved reasoning and set a new state-of-the-art for models of their sizes.

Today's release includes the first two Llama 3

Open source is the most transparent, the most collaborative, and the most secure approach to AI.

GPT-4 and other LLMs seem more like sophisticated parrots with large memories than truly creative humans. These models must be properly validated with out-of-sample data. However, none of that was done with GPT-4 because the model would need to be more focused and not appear as