Tuesday, February 11, 2025

UDLCO CRH: Building Elon's everything app with ideas ranging from practical in time to impractical futuristics

Elon Musk's innovations in Summative assessment:

Summary:

The conversation revolves around AI, healthcare, and innovation, with a focus on Elon Musk's ideas and their potential applications. The discussion touches on topics such as:

- AI in healthcare and medical education
- Data capture and processing in healthcare
- Clinical informatics and its importance in India
- Patient capital and its role in data-driven healthcare
- The potential for AI to structure unstructured data in healthcare

Key Words


- AI
- Healthcare
- Innovation
- Elon Musk
- Clinical informatics
- Patient capital
- Data capture
- Data processing






Image copyright: probably Elon or whoever wants to claim it 

Conversational Transcripts:

[16/01, 08:11] rb : Very soon he'll find a lot of AI agents sharing their code and it will be interesting when he invites them for a face to face interview

[16/01, 08:13] gm: Reality today is lots of dev automation ala code pilots etc. It is essential to know the new paradigm. It is not an either/or

[16/01, 08:16] rb: Yes as long as he's not the only human in the loop

[16/01, 08:17] So: The biggest factor is copying. I saw the same algo/project in at least 70% the resumes ...it is a publicly available algo :)

[16/01, 08:20] rb: To eliminate that possibility one needs a face to face interview to simply verify the claims made by the potential candidate. In education parlance aka summative assessment while the CV essentially reflects a formative assessment by the candidate themselves. Elon has cut short to the chase by eliminating the need for formative self assessment and straight away beginning the summative assessment albeit in an asynchronous manner, which is pretty neat




[10/02, 13:51]ak: Good afternoon all,
A doctor just asked me if I am aware of some best online courses (certificate programs) on AI for medical professionals (doctors).

Any inputs ?


[11/02, 07:24]st: Would rate iisc bangalore program as the best for doctors . Realised it after finishing course.


[11/02, 07:27]st: Learning python and advanced maths as a requirement here and their capstone project helps you make ai project with iisc faculty.


[11/02, 07:29]rb: Online or onsite?


[11/02, 07:31] st: Online but quite rigorous.  6 hours every Sunday for a year but you get amazing batch of smart people as cohort in your batch. Me and dr p were batch 7. Now it's 9th going and they keep adding new things in course. They have onsite visits and mentoring


[11/02, 07:35] st: https://otoscanai.com/ i was able to make this ai based classifier after the course. U have to put images of ear canal/ drum and it will give diagnosis . Useful for non wnt surgeon's and patients for home based diagnosis


[11/02, 07:37] st: Would be happy if any one tells me how to generate revenue from it 🤔.  That iisc never teaches. Considerable  server costs with😁 deployment


[11/02, 22:19] dt: First you have to make a business case for your product. See how it solves a pain point in the current scheme of things, probable market size, probable price points from economics / business point of view. Product manufacturing and marketing expenses. Study if any alternative products are already available in the market. If yes, then see how your product is superior / inferior to it. Visualise other potential obstacles in acceptance and commercialization of your product.

Based on the above, start working on how much funding you will need to manufacture the product. Licences required for the same etc

You will need a technical expert to figure out exactly the technical requirements for manufacturing the product 

Have you patented your product as yet or not? What stage are you in your product development cycle which ranges from concepts to readiness for commercialisation roll out.

After this with a good financial advisor make a financial plan of the business atleast for 5 yrs.

Once your business plan is ready, find another good financial consultant who will search and find potential investors for early stage funding for an equity stake. ( You can also decide how much of your own money you are willing to committ for this. Financer call it your Skin in the game to evaluate how much you are committed or serious about your business idea)

After doing this much research, if you still have any questions then you can DM me

If possible, attend some boot camps held by venture capital firms or bodies

[12/02, 07:26] rb: First step:

If it's going to be DIYWAI ear canal diagnosis then you will need to ensure adequate means of data capture for the non ent surgeon user.

Also don't make this market limited to them. Reach out with your DIYWAI to the bottom of the pyramid (all users with a ear canal and drum problem)


[12/02, 07:55]rb: Second step:

Team up with a generalist healthcare professional who can integrate all other disciplines of healthcare to build their data capture portals (various other canals and drums in the body, you get the drift) to integrate with your ear portal to build the entire healthcare data capture portal for every human user to trade off their data for a solution to their pressing healthcare requirements.

Step 3:

Team up with every engineering and science professional globally to transform your healthcare portal into an everything portal (after all essentially healthcare is everything and even now engineers and scientists are it's real developers and doctors are glorified retailers) and finally submit to Elon for his "everything app" competition because by now he's realised that tamatar data will not take him anywhere except a chaotic hive mind and he needs not only users mental health requirements but also their connected bodily requirements to create his ultimate human requirements engineering app to fit into his current neuralink and mars jigsaw!

[12/02, 07:55] dt: Health / Clinical informatics is not used in HMIS/EHRs in India. The importance of the use of clinical informatics is not very well appreciated in India. Medical professionals who are getting interested/excited about AI in health must also start taking interest in clinical informatics as this will significantly improve outcomes of their AI applications

[12/02, 08:00]ss: Health Data Literacy and Digital Literacy both are essential core topics, which should not be ignored


[12/02, 08:12] rb: Yes core topics to life long learning in data driven healthcare! 

Short term learning overestimates our desired outcomes while long term learning forces us to underestimate future returns/outcomes and one needs "true patient capital" at the right bottom end of the graph here (wrongly labeled philanthropy) to be able to achieve true data driven evidence based healthcare.



[12/02, 08:13]rb: Bottom-line:

Data capture is the more important and currently neglected step rather than AI driven data processing, which may have become easier minus the hallucinations


[12/02, 08:23] dt: Use of Clinical informatics provides structured data for AI processing hence quality of AI outputs is better.

In India, we record and store our clinical information in a semi structured & unstructured way


[12/02, 08:27] rb: We can also train AI to structure these gold mine of unstructured data using a gold amalgamation smelter model?


Thematic Analysis

The conversation can be categorized into several themes:

1. *AI in Healthcare*: The discussion highlights the potential of AI in healthcare, including its applications in medical education and data-driven decision-making.
2. *Data Capture and Processing*: The importance of data capture and processing in healthcare is emphasized, with a focus on the need for structured data and the potential for AI to structure unstructured data.
3. *Clinical Informatics*: The conversation stresses the importance of clinical informatics in India, highlighting its potential to improve healthcare outcomes and facilitate AI-driven decision-making.
4. *Innovation and Entrepreneurship*: The discussion touches on the potential for innovation and entrepreneurship in healthcare, including the role of patient capital in supporting data-driven healthcare initiatives.
5. *Elon Musk and Innovation*: Elon Musk's ideas and innovations are referenced throughout the conversation, highlighting his influence on the discussion around AI, healthcare, and innovation.

Posted by at No comments:

Sunday, February 2, 2025

UDLCO: Populist write ups on Deepseek's strengths and original author attribution challenges in forwarded social media publications

Summary: 


The conversations highlight the potential disruption of the AI industry and the importance of simplifying complex technologies and while doing so a text is shared as a viral tweet about DeepSeek's AI innovation approach that reduces training costs and GPU requirements, making AI more accessible. 
The subsequent conversations also demonstrate that original author attribution is a challenge in forwarded social media content. This raises concerns about intellectual property, credibility, and misinformation. Clear attribution and source verification are essential to address these issues.

Key Words

1. DeepSeek
2. Nvidia
3. AI innovations
4. Training costs
5. GPU requirements
6. Disruption
7. Simplification
8. Accessibility
9.Misattribution IP
10. Information diffusion 

Conversational Transcripts:

[28/01, 10:06] : forwarded

Finally had a chance to dig into DeepSeek’s … 

*Let me break down why DeepSeek's AI innovations are blowing people's minds (and possibly threatening Nvidia's $2T market cap) in simple terms...*

First, some context: Right now, training top AI models is INSANELY expensive. OpenAI, Anthropic, etc. spend $100M+ just on compute. They need massive data centers with thousands of $40K GPUs. It's like needing a whole power plant to run a factory.

*DeepSeek just showed up and said "LOL what if we did this for $5M instead?" And they didn't just talk - they actually DID it.* Their models match or beat GPT-4 and Claude on many tasks. The AI world is (as my teenagers say) shook.

How? They rethought everything from the ground up. Traditional AI is like writing every number with 32 decimal places. DeepSeek was like "what if we just used 8? It's still accurate enough!" Boom - 75% less memory needed.

Then there's their "multi-token" system. Normal AI reads like a first-grader: "The... cat... sat..." DeepSeek reads in whole phrases at once. 2x faster, 90% as accurate. When you're processing billions of words, this MATTERS.

But here's the really clever bit: They built an "expert system." Instead of one massive AI trying to know everything (like having one person be a doctor, lawyer, AND engineer), they have specialized experts that only wake up when needed.

Traditional models? All 1.8 trillion parameters active ALL THE TIME. DeepSeek? 671B total but only 37B active at once. It's like having a huge team but only calling in the experts you actually need for each task.

The results are mind-blowing:
- Training cost: $100M → $5M
- GPUs needed: 100,000 → 2,000
- API costs: 95% cheaper
- Can run on gaming GPUs instead of data center hardware

"But wait," you might say, "there must be a catch!" That's the wild part - it's all open source. Anyone can check their work. The code is public. The technical papers explain everything. It's not magic, just incredibly clever engineering.

*Why does this matter? Because it breaks the model of "only huge tech companies can play in AI." You don't need a billion-dollar data center anymore. A few good GPUs might do it.*

*For Nvidia, this is scary. Their entire business model is built on selling super expensive GPUs with 90% margins. If everyone can suddenly do AI with regular gaming GPUs... well, you see the problem.*

And here's the kicker: DeepSeek did this with a team of <200 people. Meanwhile, Meta has teams where the compensation alone exceeds DeepSeek's entire training budget... and their models aren't as good.

*This is a classic disruption story:* Incumbents optimize existing processes, while disruptors rethink the fundamental approach. DeepSeek asked "what if we just did this smarter instead of throwing more hardware at it?"

*The implications are huge:*
- AI development becomes more accessible
- Competition increases dramatically
- The "moats" of big tech companies look more like puddles
- Hardware requirements (and costs) plummet

Of course, giants like OpenAI and Anthropic won't stand still. They're probably already implementing these innovations. But the efficiency genie is out of the bottle - there's no going back to the "just throw more GPUs at it" approach.

Final thought: *This feels like one of those moments we'll look back on as an inflection point. Like when PCs made mainframes less relevant, or when cloud computing changed everything.*

*AI is about to become a lot more accessible, and a lot less expensive. The question isn't if this will disrupt the current players, but how fast?* One of the reason of market fall across the globe


[28/01, 10:09] snpc: The US businesses always suffered with this problem of not simplifying...especially if it is against their business model. Intel delayed the dual core because they were worried about cannibalizing the sales. I am sure big tech knew this all along.


[28/01, 10:26] +91: Very well summarized 👍

[28/01, 10:35] +91: Could you pl share original article url?

[28/01, 11:13] : I've asked the person who shared it with me but not sure if he would be able to get it though.

[28/01, 11:14] +91: Indeed the person who actually wrote this piece has really done a great job in simplifying things.

As per google apparently the author works in Shanghai👇



[28/01, 12:49] snpc: A nice mindmap of the Deepseek paper Source https://pbs.twimg.com/media/GiWdPsNWkAAanXx?format=jpg&name=4096x4096


[28/01, 12:52] sd: It’s not an original article. It was a tweet yesterday which went viral.

[28/01, 12:59] rb: Wasn't the tweet originally written by Jeremy? 🙂

[28/01, 13:33] sd: I don’t think so

[28/01, 13:34] sd: I recommended reading these two to understand: 




[28/01, 13:37] sd: I strongly recommend not to rely on Google for these lol 

I checked on twitter…

Jeremy had just reposted.



Inputs regarding the theme of "original author attribution challenges in forwarded social media publications" in the conversational transcripts shared:

Challenges in Original Author Attribution

1. *Lack of clear attribution*: The conversational transcripts show that the original author of the content is not clearly attributed. The text is forwarded without mentioning the original author's name or handle.

2. *Information diffusion*: The content is shared and diffused through social media platforms, making it difficult to track the original source.

3. *Loss of context*: The forwarded content may lose its original context, making it challenging to understand the author's intentions or the content's relevance.

4. *Misattribution*: The conversational transcripts suggest that the original author's identity may be misattributed or unknown, leading to potential miscrediting or misrepresentation.

Implications

1. *Intellectual property concerns*: The lack of clear attribution raises concerns about intellectual property rights and the potential for plagiarism.

2. *Misinformation and disinformation*: The diffusion of unattributed content can contribute to the spread of misinformation and disinformation.

3. *Credibility and trust*: The inability to verify the original author's identity can erode credibility and trust in the content and its sources.

4. *Academic and professional integrity*: In academic and professional settings, proper attribution is essential for maintaining integrity and credibility.

Possible Solutions

1. *Clear attribution*: Encourage users to provide clear attribution when sharing content, including the original author's name, handle, or publication source.

2. *Source verification*: Implement mechanisms for verifying the original source of content, such as fact-checking or reverse image search.

3. *Content tracking*: Develop tools for tracking the diffusion of content across social media platforms, enabling the identification of original sources.

4. *Education and awareness*: Promote education and awareness about the importance of proper attribution, intellectual property rights, and academic integrity.

Thematic Analysis with Coding on populist social media viral opinion around deepseeks approach:

1. *Disruption of Traditional AI Models* (DTAM): DeepSeek's innovative approach challenges the traditional AI model, reducing costs and increasing accessibility.

2. *Simplification of Complex Technologies* (SCT): The importance of simplifying complex technologies to make them more accessible and user-friendly.

3. *Accessibility and Democratization of AI* (ADAI): DeepSeek's approach makes AI more accessible to smaller companies and individuals, democratizing the technology.

4. *Market Disruption and Competition* (MDC): The potential disruption of the AI industry and the impact on market leaders like Nvidia.

Categorization

1. *Technology*: AI innovations, GPU requirements, training costs
2. *Business*: Market disruption, competition, accessibility
3. *Society*: Democratization of AI, simplification of complex technologies

Learning Points

1. *Innovation can come from unexpected places*: DeepSeek's approach challenges traditional AI models, demonstrating the potential for innovation from smaller companies.

2. *Simplification is key*: Simplifying complex technologies can make them more accessible and user-friendly.

3. *Disruption can lead to growth*: The potential disruption of the AI industry can lead to growth and innovation, as companies adapt to new challenges.

4. *Accessibility is crucial*: Making AI more accessible to smaller companies and individuals can democratize the technology and lead to new applications and innovations.




Posted by at No comments:
Subscribe to: Posts (Atom)