AI health progress

Investment Analysis Clubs Macro Economic Trends and Risks
1

I’ve been watching the “AI takes over Patient Healthcare” space for over a year, ever since OpenAI revealed ChatGPT.

These comments are neither investment nor healthcare advice. Just my observations.

I expect that institutional lobbists push back to be quite strong in resisting AI replacing Doctors, MD, specialists, etc.
We’ve known for more than a decade that computer n software analysis is MUCH more efficient than human analysis of X-rays, MRIs, CAT scan, etc.

Google, Amazon, OpenAI are offering their versions of AI assisted healthcare.

Here are some B2C ideas:

OpenAI offers ChatGPT Health.
https://openai.com/index/introducing-chatgpt-health/
OpenAI was a first mover effort to capture the LLM AI healthcare space, last Fall.

[ Superpower. Private?, tiny company.

Sketchy?.. LOLOL.
Was the second entry to try for 1st mover.]

Google offers it’s “health” AI assisted by Gemini AI. I’ve not seen any real “comprehensive one stop shop” type version from Google.
(Personally, Google already knows all about me, so I’m waiting for Google’s AI LLM app.)

Amazon offers Amazon Prime Health.

Apple via its Watch n other wearables.
And, with its huge dedicated fan base, is an easy “winner”. (* See Edit below.)

MSFT surely has plans in this space?
META, ditto.

NVDA n PLTR have partnerships with “everybody for everything”.

These Mag7s have a lot of lobbying power, too, to counter the Big Medicine lobbists.
But, they are megacaps n their AI health side hustles are unlikely to move their needle.

B2B: these companies offer tech to B2C businesses
Oura n a plethora of other wearable health monitoring tech n apps.

I sorta expect Mark Cuban n CostPlusDrugs to put its toes in the water.

I’m watching for potential investments.

Here’s one such that crossed my YT feed this morning:

ONMD is one such.
It is a WEAK company, according to the YT. (Dilution risk!)

Penny stock; 1.10$/share.
Looks like it went public in 2021 via SPAC?
Historically, most SPACs fail.
HQ in Minnesota.

(OneMedNet Corporation, a healthcare software company, provides digital medical image management, exchange, and sharing solutions).

This has huge promise - of being acquired. IMO. LOLOL.

https://finance.yahoo.com/quote/ONMD/profile/

Butterfly IQ ultrasound.

Nano-x imaging and analysis.

…
Given the “unhappiness” many folks have with running back and forth to a “clinic” just to “talk”, I think there will be a huge move to “telehealth”.

Hospitals, Dr offices, etc are FORCING clients to use AI assistants for contact.
As more n more of their clients age/die out, I expect more clients will shift their healthcare interests to fully Al/LLM telehealth.

:medical_symbol::syringe::pill::x_ray::man_in_lotus_position:
:man_in_lotus_position::man_walking::man_lifting_weights:. :money_mouth_face:
ralph observes. No advice.

Edit: Nate makes the case for Apple.

6 Likes
2

The AI goldrush is on.

My software in game development is in a massive, very effective debugging phase. it is a must. When In the Microsoft Developer Program, I entered my game with them for review. It can not be buggy. It won’t be, but I by myself would get nowhere. Even a partial success would be an invitation for endless tinkering.

1 Like
3

Last night at dinner, BIL (whose job involves building super computers) was talking about at work using AI to help rate proposals and generally get information. He said they were using a paid version of AI (can’t remember which one now) and I’ve read several articles talking about paid versions being “better” than the free version the majority of people are currently being exposed to. Long story short, he said the AI was very comical, in a bad way. Would give it parameters to grade various proposals and it would often ignore the parameters. No mater how he rephrased things. In its current form, he rarely uses it.

So if BIL is having issues, and it is his job/wheelhouse, I don’t see it being effect in medicine enough to replace MDs, nurses, etc. in its current form.

Always consider the Gell-Mann Amensia effect.

3 Likes
4

@JLC

I have used ChatGPT, Gemini, and Claude. Chat is making the most commercial success, but it is a racket. While extremely fast, it is inaccurate. It seems the easiest to use, but it gets bogged down later. That is when you realize you need to pay for it. Since TPTB are aimed at using Chat, it might be that once you pay for Chat, you still get the inaccuracy. Chat does less with the project structure.

In coding, accuracy is everything. The program won’t run or won’t give you the result you want otherwise.

Gemini is good, but not made primarily for coding. Gemini is for the overview of a project. But Gemini does not do well with details. So if I were to restructure my code with Gemini, the new structure with all the new options is not grounded in what I need. I discovered that as I noticed none of the options were relevant.

Claude is excellent. Reality bites. Claude limits the user dramatically. I think Claude wants to run in the black. I now pay $200 per month for Claude. That is with a 50% discount for this month. I can only hope Anthropic keeps the rate the same next month. Claude is so good, I might be done next month before the bill comes due May 22. If not I will opt for the $100 Pro plan. I would be just finishing up.

Claude coordinates accurate coding with excellent structural forms.

It is as if all of them have the same levers to push. You pick your poison.

Back in college logic, I noticed that the fallacy of parts or whatever it was called was the most difficult to overcome. The Japanese army is made up of small men, there for the Japanese army is small. Claude succeeds with flying colors.

1 Like
5

I concur!
I use free versions of Gemini, Claude, Perplexity, ChatGPT. All of them, to date, require critical thinking in order to recognize that “something ain’t right” n then reword or revise the prompt.
So far, I’ve been able to resolve issues to my liking. :slightly_smiling_face:
The LLMs are steadily improving.

My goal, in following this space, is to track progress and evolution of the tech, and to identify potential investments.
IMO, there is, so far, not a single entity that covers/provides the entire stack from data collection to data storage to data analysis to data sharing between “silos” to “comprehensive data analysis and summarization” to “end healthcare delivery to the end user”.

I suspect that the health care industry will successfully lobby and maintain the “an accredited/degreed MD must verify/approve the diagnosis and delivery of the health care medium”, at least for the foreseeable future. But that’s a political discussion.

Never the less, I expect AI use and AI assisted telehealth to increase, if for no other reason than “it reduces cost and allows higher margins and revenues to the provider”.

I’ve used free Grok, but it’s onerous and often seems to glitch.

:slightly_smiling_face:
ralph

1 Like
6

One investing website I visit has recently launched an AI tool. I’ve played around with it with little success. Just trying to get it to gather a list of stocks according to Graham Value was frustrating. Would feed it the parameters but then it was unable to find websites with the data, even simple things like EPS, etc.

We are far from the Borg and/or Terminator.

7

Look up the paper “Mirage: The Illusion of Visual Understanding”

1 Like
8

Absolutely!

Data is notoriously variable from silo to silo and even from data lake to data lake.
EVERY ‘company’ has it’s system for collecting and storing data/info.
This leads to different format, different descriptors n definitions, differences in all kinds of things. Structured, semi-structured, unstructured, etc.

SNOW, DataBricks, MSFT n AWS have products that address this? These companies aggregate the siloed ‘raw data’ to data lakes, and into ‘useable’ formats.
This is one level of the ‘stack’ I mentioned above.

EVERY ‘company’ wants to control access to ‘their’ data. Today, we are told ‘data is the new gold’. EVERY ‘company’ wants to be PAID for ‘their’ data.
Agreements must be made between companies such that the data owners get paid, while also maintaining privacy and security of the data.
This is sort of a ‘middle layer’ of the stack I mentioned above.
ONMD fits into this layer.

(Disclosure. I now own some ONMD.)

from Yahoo Finance:
(OneMedNet Corporation, a healthcare software company, provides digital medical image management, exchange, and sharing solutions in the Americas, Europe, and the Middle East. The company offers iRWD, which provides regulatory grade imaging and clinical data in the pharmaceutical, device manufacturing, contract research organizations, and artificial intelligence markets markets; and BEAM, a medical imaging exchange platform between hospital/healthcare systems, imaging centers, physicians, and patients. OneMedNet Corporation is headquartered in Eden Prairie, Minnesota.)

The ‘top layer’ of the stack is where many disparate pieces of info are aggregated into a diagnosis and a treatment plan.
This is where I see the LLMs I mentioned in the OP (OpenAI, Google, Amazon, etc.)
And is where the end user patient will be ‘serviced’.
So far… I’m not seeing any company offering this. IMO, it’s coming. The first ‘offerings’ will be less effective. But, as time passes, will become much better.
IMO, it will require a massive company such as Google to adequately cover all the security and privacy aspects.

Google debuted some ‘Google Finance’ LLM thing … last Fall?

I’ve NOT used the Google Finance LLM.

Schwab offers Think or Swim (ToS), which is a COMPREHENSIVE tool that predates AI n LLMs. It’s COMPLEX. I have played around with it, but I like/prefer ‘simple’. It offers ‘tracking and alerts’. It might fit your FI n Graham Value needs? I do NOT know if ToS incorporates LLM AI functionality.
I suspect Schwab/ToS either already offers it, or ‘has plans’.

:slight_smile:
ralph

9

Artificial Assistance - in today’s version is an efficiency impact for adept users.

For most users (and many uses!) these algorithms are helpful to the LEAST capable and LEAST knowledgeable.

NOTE: . it is NOT exclusive to be adept, but not capable OR knowledgeable

. . . . . . . or

. . . . . . . to be CAPABLE and/or KNOWLEDGABLE but clumsy at AI.

AI is here for the LEAST capable today, and offers large benefits. This group knows little, understands little, but gets more benefit than harm with current tools.

This will not be the case later.

For all of us on the pointy end, ADEPT AI use is a benefit also. It is our knowledge and capability which enables this.

AI also provides ample ground for humor, amusement and concern.

2 Likes
10

So did all the rights holders for the data these “companies” trained on.

2 Likes
11

There is the danger in using AI. They don’t know enough to know that the AI is wrong or not exactly correct in its answer. Kind of like people saying doctors can be replaced with PAs/NPs/etc and my response has been what they don’t know can kill you.*

* I’ve worked with many “physician extenders” in my career and the majority are good and capable. But the training and knowledge base is drastically different.

2 Likes
12

I agree completely. But AI is helping them. Errors and mistakes related to their use are shared between the individual, the program, and, for official workstreams, the company that placed the AI into the process.

This is another variation about ignorance being bliss… until it isn’t. Grave concerns, indeed.

1 Like
13

I have academically coded in 5 languages decades ago. I know Photoshop, AutoCAD, and Blender. I have used PS and Blender extensively. Now I am using visual scripting with Unreal Engine. Most people should just head to the exits first. LOL

I am not even mildly dangerous on my own in the coding world. But with AI, I have a massive edge. I can create. So I can make things that otherwise would never fly. That is why this is a gold rush. Some of us are getting opportunities we never would have had.

1 Like