Benjamin Wallsten (@1humanatwork) header
Benjamin Wallsten (@1humanatwork) profile picture

Founder, CEO @createwithapril | EP w/ Del Eswar at First Law Motion Pictures | @AntlerGlobal alum | #CreatewithApril

NYC
Joined February 2012
Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)Photo by Benjamin Wallsten (@1humanatwork)
Benjamin Wallsten retweeted
In a remarkable twist of fate, a group of divers recently stumbled upon an extraordinary archeological discovery hidden beneath the serene waters of Lake Lucerne in Switzerland. The accidental unearthing of a 3,000-year-old sunken village has sent shockwaves through the historical community and has provided a profound insight into the region's ancient past. Diving deep into the depths of Lake Lucerne, archeologists found themselves immersed in a world frozen in time. Beneath a thick layer of sediment and mud, they unearthed remnants of a bygone era. Shards of pottery, reminiscent of an intricate civilization, painted a vivid picture of a flourishing society that once thrived in this picturesque region. At the heart of this submerged village lay the remains of wooden pilings, a poignant reminder of the homes that once stood tall upon them. These wooden structures, now preserved by the lake's protective embrace, served as the foundations for houses that were inhabited over three millennia ago. It is believed that these dwellings were constructed during the Bronze Age, a period that spanned from around 3300 to 1200 BCE. The discovery of this sunken village has rewritten the history of Lake Lucerne and its surroundings. Until now, it was widely believed that the region was sparsely populated during ancient times, with little evidence of significant settlements. However, the find demonstrates that a vibrant community existed here long before modern civilization took root. This revelation has opened up a wealth of questions for historians and archeologists alike. Who were the people that once called this submerged village their home? What was their way of life? What led to the eventual submergence of the settlement? These inquiries may offer valuable insights into the societal dynamics and environmental changes that shaped the region during the Bronze Age. The implications of this discovery extend far beyond the confines of Lake Lucerne. It adds another puzzle piece to the larger tapestry of human history, offering a glimpse into the lives of our ancient ancestors. The artifacts recovered from the sunken village provide a tangible link to the past, allowing us to gain a deeper understanding of our shared heritage. The underwater archeologists responsible for this remarkable find are working meticulously to piece together the fragments of the past. They carefully document and analyze each artifact, employing cutting-edge techniques to ensure their preservation and to extract every morsel of information they hold. The diligent efforts of these experts will help unravel the mysteries of the submerged village and shed light on a previously unknown chapter of human civilization. As news of the discovery spreads, researchers and historians from around the world are eagerly awaiting the revelations that will emerge from this extraordinary find. The submerged village in Lake Lucerne stands as a testament to the resilience of the human spirit and the inherent curiosity that drives us to uncover our past. With every dive into the depths of Lake Lucerne, the secrets of this sunken village are slowly coming to the surface. As the story of this ancient community unfolds, it serves as a reminder that there are still hidden wonders waiting to be unearthed, even in the most unexpected places. The accidental discovery of this 3,000-year-old sunken village in Switzerland ignites our imagination and fuels our relentless pursuit of knowledge, inviting us to delve further into the annals of history and explore the vast depths of human civilization.
408
2,779
197
28,459
13,115,583
Benjamin Wallsten retweeted
#Config2023 launches bridge the gap between design and development, all in Figma. → Dev Mode, a new space for developers → Variables → Advanced prototyping → Auto layout updates → Font picker → File browser redesign Plus, we previewed the future of Figma with AI and announced the acquisition of @diagram. figma.com/whats-new
229
2,109
940
8,120
1,338,030
Benjamin Wallsten retweeted
AI-generated QR codes using ControlNet are insane. This is going to be increasingly common in ads in the near future. These examples blew my mind (try scanning them): 1. Ancient Village
379
3,383
510
22,147
9,951,931
Excited to announce 🛠️Neuralangelo🛠️! Neuralangelo turns images into 3D meshes with extremely high fidelity and scales up in large environments. Come and chat with us during #CVPR2023, or check out our project page: research.nvidia.com/labs/dir… More details in thread below (1/4)
9
33
5
184
27,867
Benjamin Wallsten retweeted
To give you a sense of how fast AI for 3D modeling is advancing: the field went from the left (original NeRF-reconstructed mesh) to right (Neuralangelo from NVIDIA) in 3 years. Transporting reality into high-fidelity simulation is no longer a pipe dream.
33
181
15
1,261
426,943
Benjamin Wallsten retweeted
Reading OpenAI's latest paper: "Let's Verify Step by Step". The idea is so simple that it fits in one tweet: For challenging step-by-step problems, give reward at each step, instead of a single reward at the end. Basically, dense reward signal > sparse. The Process Reward Model ("PRM") is able to pick solutions for the difficult "MATH" benchmark much better than the Outcome Reward Model ("ORM"). The obvious next step is to finetune GPT-4 with PRM, which the paper hasn't done yet. A caveat is that PRM requires a lot more human labeling. OpenAI releases their human feedback dataset: 800K step-level labels across 75K solutions to 12K MATH problems.
36
209
27
1,368
304,435
Benjamin Wallsten retweeted
Well-said! Voyager is an experiment on replacing the agent reinforcement learning stack with text I/O, database, and code interpreter. Code execution is a powerful means to self-refine & drive exploration continually. The cornerstone is a good coding LLM. GPT-3.5 is noticeably worse than GPT-4 in our systematic study.
What if you used LLMs for… all parts of RL? GPT-4 as the Actor and Critic GPT-4 as the Bellman Update GPT-4 as the reward function Vector DB as the replay buffer GPT-4 as the importance sampling logic
8
28
2
185
86,578
Benjamin Wallsten retweeted
Imagine being in AI and NOT being followed by @Scobleizer If you’re out there, he will find you.
43
8
5
127
211,222
Benjamin Wallsten retweeted
Evidence of Meaning in Language Models -Train LM w/ next token prediction on corpus of programs -Precisely defining concepts of correctness, semantics -Model learns to generate correct programs shorter & semantically different than those in its training arxiv.org/abs/2305.11169
5
85
3
299
30,253
Benjamin Wallsten retweeted
I'm hunting down the best ChatGPT Plugins with real use cases. I've found a winner in @Zapier. Most of the people that follow me are here for AI Autonomous Agents, so I know you guys are going to love what’s possible with Zapier.
56
182
14
1,790
952,938
Benjamin Wallsten retweeted
We just released our work on robot soccer. I've been working on this for quite some time with my amazing colleagues at DeepMind. It's exciting how deep RL can produce such beautiful behaviors with low-cost robots. Full paper is available at arxiv.org/abs/2304.13653 Enjoy!
91
551
256
2,707
759,251
I’ve heard confusion on why there are so many foundation models and this is the key point. We will need differentiated options across cost, latency, task, reliability, open source, and domain sensitivity. Eventually, model chaining is the norm.
Asking “what’s the best model?” is like asking “what’s the best code base?”, there’s no good generic answer, it depends highly on your use case, constraints, and current tech stack! Ultimately the best one will be the one you can train, optimize, control and customize the fastest
1
4
1
15
3,913
Benjamin Wallsten retweeted
Introducing 100K Context Windows! We’ve expanded Claude’s context window to 100,000 tokens of text, corresponding to around 75K words. Submit hundreds of pages of materials for Claude to digest and analyze. Conversations with Claude can go on for hours or days.
202
973
615
5,052
2,475,826
Benjamin Wallsten retweeted
here is an alternative path for society: ignore the culture war. ignore the attention war. make safe agi. make fusion. make people smarter and healthier. make 20 other things of that magnitude.   start radical growth, inclusivity, and optimism.   expand throughout the universe.
1,602
2,536
781
19,339
3,364,759
Benjamin Wallsten retweeted
Cooking up some very cool things for our community of real estate professionals, startups, galleries, and content creators. Stay tuned 😉
2
2
96
Benjamin Wallsten retweeted
This 🤯 is a very big 🤯 I have access to the new GPT Code Interpreter. I uploaded an XLS file, no context: "Can you do visualizations & descriptive analyses to help me understand the data? "Can you try regressions and look for patterns?" "Can you run regression diagnostics?"
133
1,025
260
5,640
2,068,333
Benjamin Wallsten retweeted
This is why I encourage people to just play with these tools. There is no “right” way. I just filmed myself walking up to the AMC 16 and threw it into Luma. I did not do it “right”. And I get these ethereal effects, like you are seeing bits of a separate dimension coming though. Now to run through Kaiber or Controlnet to give it a style.
48
79
15
709
375,861
Benjamin Wallsten retweeted
Replying to @AndrewYNg @OpenAI
Looks good. I found this openai article for reliable prompts very useful. github.com/openai/openai-coo…
4
32
3
251
25,944
Benjamin Wallsten retweeted
1/ Thrilled to announce: Our new course ChatGPT Prompt Engineering for Developers, created together with @OpenAI, is available now for free! Access it here: deeplearning.ai/short-course…
484
4,331
661
16,779
4,133,178
Benjamin Wallsten retweeted
Announcing StableVicuna, the AI World’s First Open Source RLHF LLM Chatbot! @CarperAI (in an effort led by @phungvanduy1) is releasing an RLHF-trained version of Vicuna-13B Learn more → stability.ai/blog/stablevicu…
19
235
29
784
333,813