Param Thakkar

365 posts

Param Thakkar banner
Param Thakkar

Param Thakkar

@Param23072004

Playing with Reinforcement Learning, World Models + GPUs, CS undergrad @ Veermata Jijabai Technological Institute, Mumbai,Independent contractor @transformerlab

Katılım Ağustos 2023
1.1K Takip Edilen60 Takipçiler
Param Thakkar
Param Thakkar@Param23072004·
@thdxr This is much more common on LinkedIn. Just take any post longer than 75 words and put it on pangram and you will find this. Even the commenters have their comments AI generated 🙂
English
0
0
2
1.3K
dax
dax@thdxr·
the world is so screwed
dax tweet mediadax tweet media
English
33
1
239
36.1K
Param Thakkar retweetledi
Awni Hannun
Awni Hannun@awnihannun·
Because of AI people are starting to value experience in a domain more than they used to. It feels short sighted. - Many (if not most) of the best ideas come from junior people with fresh eyes or senior people cross-pollinating across domains - Thanks to AI you can learn much faster
English
15
11
139
10.5K
Param Thakkar
Param Thakkar@Param23072004·
@CodeByNZ True. Much better than months of vague updates leading to rejection. If companies don’t mind wasting a candidate’s time, they shouldn’t mind being direct. It saves time for both the company as well as the candidate and clarifies the company's own priorities.
English
0
0
2
811
Alaya Studio
Alaya Studio@alayastd·
𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗪𝗼𝗿𝗹𝗱 𝗥𝗲𝗻𝗱𝗲𝗿𝗲𝗿 A new toolkit, dataset, and baseline for world-scale rendering: - collect G-buffers from AAA games - scale data for rendering complex world scenes - improve rendering performance - enable game effect editing
English
9
28
212
15.9K
Param Thakkar retweetledi
Transformer Lab
Transformer Lab@transformerlab·
Transformer Lab “tasks” let you run complex ML workflows with a single click. Import a task from the Task Gallery, configure your parameters, and run. Each task packages all setup and dependencies so you skip the troubleshooting. 🎬 Our Wan2.1 text-to-video task is a great example. Running it normally requires significant setup work. As a Transformer Lab task, you one-click import it, type a prompt and you're generating video. 🧪 The Task Gallery covers training, fine-tuning, evaluation and more. 🛠️ Create your own tasks and share them with your team. 💻 Runs on your local machine, an on-prem cluster or a cloud provider like @runpod. Open source and free. Try it out: lab.cloud
English
0
1
4
97
Param Thakkar retweetledi
Naval
Naval@naval·
Vibe coding is more addictive than any video game ever made (if you know what you want to build).
English
1.7K
2.8K
29.1K
1.6M
Param Thakkar
Param Thakkar@Param23072004·
@rosstaylor90 @GenReasoning Also @rosstaylor90, I am working on a couple of projects in RL and Mech Interp research and side projects in RL + World Models. One is currently public 2 others are private repos but would love to chat about them if useful. Thanks!
English
0
0
1
100
Ross Taylor
Ross Taylor@rosstaylor90·
Post-launch reminder that we’re hiring for several roles at @GenReasoning in London. Few things about our culture: - Clipped hype/delivery ratios - Cares about openness and AI safety - Many legends from the open model space - Rigorous and doesn’t cut corners - Knows how to exhaust a baseline before trying new things - Kind, empathetic and collaborative - Team is generally averse to being on X (except me…) and prefers doing the work DMs open for fun roles in infra/product/research 🙂.
English
9
6
104
7.7K
Param Thakkar
Param Thakkar@Param23072004·
PS: The foundation and focus of TorchWM and my broader work on World Models is deeply inspired by the ideas of @PimDeWitte and others at @gen_intuition especially the focus on using WMs for games. Would love to see WMs implemented by me in TorchWM play CyberPunk2077 soon😄
English
0
0
1
37
Christoffer Bjelke
Christoffer Bjelke@chribjel·
uhm.... so where do i report security issues with the claude code source code?
English
9
1
110
15.1K
ellen livia ᯅ 🇺🇸🇮🇩
This week in security: - LiteLLM, backdoored release exfiltrating secrets - Axios, supply chain malware via dependency - Railway, CDN caching leaked user data - OpenAI Codex, command injection via GitHub branch names - Mercor 1TB data leak - Delve, data leak + compliance risk infra is the attack surface now
English
83
513
3.2K
182.7K
Param Thakkar
Param Thakkar@Param23072004·
I think I would need larger and more GPUs to run larger models and more simulations. I have access to a 12gb vram rtx 3090 and a dgx but let's see🙂
English
0
0
2
33
Param Thakkar
Param Thakkar@Param23072004·
All the world models are now in place for TorchWM! I will keep adding newer implementations to it. There are many more in the PRs section. The next target would be using these world models as simulators to generate environments purely from imagination.
Param Thakkar tweet media
English
0
0
2
43
Param Thakkar
Param Thakkar@Param23072004·
All you need is a research mindset and the willingness to do things, to work hard on them. Even the most successful AI companies today have a CEO who is a college/high school dropout let alone a PhD. An idea and willingness to work hard and bring it to reality is all you need
Aakash Gupta@aakashgupta

Alec Radford has 190,000+ citations, no PhD, no master's degree, and 34,000 Twitter followers. Sam Altman called him an Einstein-level genius. Wired compared his role at OpenAI to Larry Page inventing PageRank. He still prototyped most of his work in Jupyter Notebooks. The resume is staggering when you list it out. GPT-1: first author. GPT-2: first author. CLIP: primary author. Whisper: co-author. DALL-E: co-author. DCGAN: co-author. Contributing researcher on GPT-3, GPT-4, and DALL-E 2/3. Multiple U.S. patents owned by OpenAI. He joined in 2016 with a bachelor's degree from Olin College, a school founded in 1997 with fewer than 400 students. His first experiment at OpenAI was training a language model on 2 billion Reddit comments. It failed. But the organization gave him room to keep going. Two years later he built GPT-1 alone, based on what colleagues described as pure technical intuition. He couldn't fully explain how it worked at the time. He just knew it would. At NeurIPS 2024, Ilya Sutskever singled out two people as responsible for the pre-training era: Radford and Dario Amodei. All four original authors of the GPT paper have since left OpenAI. Radford left in December 2024 to do independent research. His last tweet was in May 2021. A reply explaining why GPT-1's layer width was set to 768. The person who built the foundation of a $300B+ industry communicates less publicly than most interns. That ratio between impact and visibility is the strongest signal of who actually does the work versus who narrates it.

English
0
0
1
69