Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
ChatGPT 4Training data size is 570 GbNumber of parameters is 1.7 trillion assuming parameter is a simple float(4 bytes) value total size is 4 * 1.7 trillion bytes = 6800 GbParameters size is much larger than training data size. Is this not classic case of overfitting ? How is this a "intelligent system" and why is there so much hype around? is it coz psychologically convinces us that its capable of almost human like conversations?? My take on gpt4 is that they have overfit the universal dataset and obviously most queries will be subset of this universal set and response is going to psychologically satisfying irrespective of its accuracy!! And we gave a fancy terminology for this inaccuracy ..we call it "model hallucination" and the marketers market it as a creative thing being done by some AI system Welcome your thoughts!!
40
8 Comments
Pawan Lamba
General Manager Data Science at Paytm
1y
- Report this comment
I believe, Deep Learning Training Theory is not very well understood.Generalisation, Model Size (Over-Parameterization) and Over-Training has See-Saw relationship in Deep Learning Theory.Some commonly established laws of ML Training theory don't apply to DL.I would mention some of the research on this in context. - Generalization in Over-Parameterized Networks or Infinite Width Networks.- Neural Tangent Kernels.- Phenomenon of Dual Descent in DL training. - Grokking Theory in DL Training
1Reaction 2Reactions
Sravan Bodapati
Principal Scientist and Sr.Manager of Applied Science - Amazon AGI Foundation Models
1y
- Report this comment
It also exhibits efficient retrieval of knowledge/memory along with memorization of the content - we value that effective retrieval from its memory based on the context, question.
1Reaction 2Reactions
Raviteja Ganti
PSE@GEP | Ex - Microsoft | BITS Pilani
1y
- Report this comment
https://www.businessinsider.in/tech/news/chatgpt-cost-a-bomb-to-make-with-openais-losses-growing-to-540-million-last-year-report-says/articleshow/100017874.cmsThis was in news recently , the computing cost is far more high. If going by your stats, this is definitely not a viable product in long term.
1Reaction
Harvey Castro, MD, MBA.
AI in Healthcare Expert, Keynote Speaker, GPT Advisor, Former-CEO Of A Healthcare company, Author. Leading the AI healthcare revolution, developing & advising on AI applications. Recovering ER doctor turned AI futurist.
1y
- Report this comment
Sharing what a Trillion is in relation to time. https://www.linkedin.com/posts/harveycastromd_ai-healthcare-chatgpthealthcare-activity-7098247970134917120-5IQF
1Reaction 2Reactions
To view or add a comment, sign in
More Relevant Posts
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
Excited to announce I’m #OpenForBusiness and providing services on LinkedIn. Check out my services page for Business Consulting, IT Consulting and Application Development.
This content isn’t available here
Access this content and more in the LinkedIn app
20
3 Comments
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
Join Me in Filing a PIL for Fair Tax Policies and Better Public ServicesDear All,The recent Indian budget has raised significant concerns:Insufficiently revised tax Brackets: Not all tax brackets were adjusted for inflation, increasing our tax burden without a corresponding increase in purchasing powerLong-Term Capital Gains Tax: The tax rate increased from 10% to 12.5%, and the indexation benefit for properties was removed, resulting in higher taxable gains espicially when the real-estate assets are olderShort-Term Capital Gains Tax: Continued taxation on short-term capital gains affects those needing to liquidate investments quickly.Despite high taxes, we face poor infrastructure, inadequate healthcare, and substandard education, unlike Western countries with similar tax rates.Why should we accept high taxes when our public services are lacking?I am planning to file a Public Interest Litigation (PIL) against the government instead of just social media ranting and need your support.What can you do?Join the Cause: Stand with me for fair tax policies and better public services.Spread the Word: Share this post and encourage others to join.Provide Expertise: Legal guidance and support would be invaluable.Together, we can demand a fair and just system. Comment or message me with your contact if you're willing to join this fight and share this post with your friends.#revisitBudget2024 #improvepublicservice #nomorepotholes #incometax #directtax #stoptaxterrorism
37
1 Comment
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
What is MLOps? MLOps (Machine Learning Operations) streamlines the development, deployment, and upkeep of ML models, fostering collaboration between data science and IT.Efficiency, scalability, reliability, collaboration, and maintenance are why MLOps matters:- Efficiency: Accelerates model deployment.- Scalability: Manages large data volumes effectively.- Reliability: Establishes sturdy models.- Collaboration: Boosts teamwork among teams.- Maintenance: Enables continuous model monitoring.To embark on your MLOps journey:1. Build a Team: Include data scientists, engineers, and IT professionals.2. Choose Tools: Opt for platforms like MLflow, Kubeflow, TFX, or Airflow.3. Automate: Simplify data pipelines and deployment processes.4. Implement CI/CD: Ensure seamless integration and deployment.5. Monitor: Keep tabs on and enhance model performance.#MLOps #MachineLearning #DataScience #IT #Collaboration
12
1 Comment
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
More often than not, I hire fullstack data scientistsThey have expertise in data engineering, analysis, machine learning, and deploymentEven the junior most member of my team brings the following on to the tableEfficiency: Handle projects end-to-endVersatility: Adaptable to various tasks and rolesCost-Effective: More economical than multiple specialistsSpeed: Accelerate development and deploymentInnovation: Drive new opportunities and optimize operations.Invest in fullstack data scientists to build scalable solutions like I do!#datascientist #ai #mlops #technology
33
3 Comments
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
Over my 17-year career, I've witnessed a jolting transformation from an employee-centric culture to one sharply focused on layoffs. About 40k employees got fired in 2024 alone in India so far!A person laid off from a company is a customer lost for many companies! Hope the companies focused on cost cutting realize it soon!
27
1 Comment
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
My friend, a seasoned data science professional with 13 years of experience as a data scientist and 4 years as a data science trainer, is seeking a new opportunity. She is aiming for a salary range of 25-30lacs and is looking in Bangalore, Hyderabad, or Kolkata. Any leads or assistance would be immensely valued. Thank you in advance! #jobsearch #datascience #opportunity
8
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
Excited to welcome Aditi Sharma to our team at Sixt R&D ! Her talent and enthusiasm are sure to make a great addition. Here's to achieving new heights together! #WelcomeAboard #TeamGrowth
11
1 Comment
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
Excited to welcome Jigyasa Singhal to our team at Sixt R&D. Her talent and enthusiasm are sure to make a great addition. Here's to achieving new heights together! #WelcomeAboard #TeamGrowth
23
2 Comments
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
If AI replaces all jobs in the future, the consequences could be dire. Without employment, people will have no money to spend. Without people spending, businesses would shrink. The result could be a severe economic recession.Do all the tech giants already forsee this problem? Will they limit the development of AI to just assistive tools? As its obvious to me that everyone will be doomed with extreme advancement of AI.
28
7 Comments
Like CommentTo view or add a comment, sign in
-
Jagat Chaitanya P.
Head of Data Products @ SIXT R&D India | Responsible AI
- Report this post
Hiring for data science positions. Having experience in building LLM & RAG is mandatory. Please message if you are interested and apply in the below linkhttps://lnkd.in/gThvJgsx
21
1 Comment
Like CommentTo view or add a comment, sign in
- 70 Posts
View Profile
FollowExplore topics
- Sales
- Marketing
- IT Services
- Business Administration
- HR Management
- Engineering
- Soft Skills
- See All
