Aanchal Agarwal

[abhishek @abhi1thakur]

The most liked models on Hugging Face in 2024 🔥

[Aanchal Agarwal @aanchaldagarwal]

Discover how Aristotle’s first principles offer a timeless approach to solving complex problems, providing clarity and direction for impactful solutions.

From Ideas to Impact: Solving Problems with Aristotle’s First Principles https://t.co/RIfFPecdjR

#FirstPrinciples

[Andrej Karpathy @karpathy]

# On the "hallucination problem"

I always struggle a bit with I'm asked about the "hallucination problem" in LLMs. Because, in some sense, hallucination is all LLMs do. They are dream machines.

We direct their dreams with prompts. The prompts start the dream, and based on the LLM's hazy recollection of its training documents, most of the time the result goes someplace useful.

It's only when the dreams go into deemed factually incorrect territory that we label it a "hallucination". It looks like a bug, but it's just the LLM doing what it always does.

At the other end of the extreme consider a search engine. It takes the prompt and just returns one of the most similar "training documents" it has in its database, verbatim. You could say that this search engine has a "creativity problem" - it will never respond with something new. An LLM is 100% dreaming and has the hallucination problem. A search engine is 0% dreaming and has the creativity problem.

All that said, I realize that what people *actually* mean is they don't want an LLM Assistant (a product like ChatGPT etc.) to hallucinate. An LLM Assistant is a lot more complex system than just the LLM itself, even if one is at the heart of it. There are many ways to mitigate hallcuinations in these systems - using Retrieval Augmented Generation (RAG) to more strongly anchor the dreams in real data through in-context learning is maybe the most common one. Disagreements between multiple samples, reflection, verification chains. Decoding uncertainty from activations. Tool use. All an active and very interesting areas of research.

TLDR I know I'm being super pedantic but the LLM has no "hallucination problem". Hallucination is not a bug, it is LLM's greatest feature. The LLM Assistant has a hallucination problem, and we should fix it.

</rant> Okay I feel much better now :)

[Daniel Paleka @dpaleka]

Watching a talk on *LLM evaluation* organized by Langchain, featuring guests from OpenAI and Anthropic. Main takeaways: (1/11)

[Aanchal Agarwal @aanchaldagarwal]

@flyspicejet Hey, the wheel chair got added. Thanks

[Aanchal Agarwal @aanchaldagarwal]

@flyspicejet Hi, need a wheelchair for my grandmother who is a senior citizen
PNR: LFM1QQ
Flight No: SG 184
Date: 18th April, 2023 that is TODAY**.

[Conor @jconorgrogan]

I dumped a live Ethereum contract into GPT-4.

In an instant, it highlighted a number of security vulnerabilities and pointed out surface areas where the contract could be exploited. It then verified a specific way I could exploit the contract

[Daniel Paleka @dpaleka]

What happened this month in AI/ML safety research. 🧵 (1/8)

[Akshat Shrivastava @Akshat_World]

First we pay 28% tax on a car (non-luxury, small).
Then we pay 70% fuel tax to drive that car.
Then we pay road tax to build roads.
Then we pay toll tax to use highways.

Welcome to the world of high taxes.

[A thread...]

[David Eagleman @davideagleman]

"Ten Unsolved Mysteries of the Brain"-Discover Magazine. I wrote this article 15 years ago. Amazingly, these are still the mysteries we face. 

https://t.co/SIfEB8VkUv

[Andrej Karpathy @karpathy]

GPT is a new kind of computer architecture that runs on text. Yes it can talk to us, but also to much of our existing software infrastructure. First via apps on top of APIs, now inside ChatGPT via plugins.
What a time right now...
https://t.co/HjeUCv3XE7

[Daniel Paleka @dpaleka]

What happened this month in AI/ML safety research. 🧵 (1/9)

[Car Dealership Guy @GuyDealership]

That's all for now.

If you found this thread helpful, RT the tweet below to share it with your audience:

[NTC @ntcpune]

"The learning that we are creating in our classrooms does not promote engagement with one's social milieu and as a result of which one develops a desire to escape and that is what is creating volatility into the system." - Dr . Ved Prakash #2ndNTC #MITWPU

[NTC @ntcpune]

"My inner voice is telling me that it's time to change profession. As a police officer, I have worked for the people of Mumbai and Maharashtra for many years, but now it's time to work for the entire country with a renewed energy." - Dr. Satya Pal Singh #2ndNTC @dr_satyapal

[NTC @ntcpune]

National Teacher's Congress has evolved as a unique platform of this kind of interdisciplinary discussions where all the teachers from all across the disciplines of Education come together. #2ndNTC #MITWPU

[NTC @ntcpune]

"The moment you start feeling your teaching and learning, i.e, education like breath, you will be a good ideal teacher" - Grp. Capt. D. P. Apte (Retd.) #2ndNTC #MITWPU

[Mansi Swaraj @Mansi0616]

#2ndNTC
#MITWPU https://t.co/TSnTTBiPYC

[Aanchal Agarwal @aanchaldagarwal]

#2ndNTC #MITWPU https://t.co/LDjhJEKSwU

[Mansi Swaraj @Mansi0616]

#2ndNTC
#MITWPU https://t.co/TSnTTBiPYC

[Aanchal Agarwal @aanchaldagarwal]

#2ndNTC #MITWPU https://t.co/QnauiNTC42

[Akshada1041 @AkshadaLawate]

#2ndNTC #MITWPU https://t.co/ALkMQY9ra6