Martineski

Significance

Adaptive agents must continually satisfy a range of distinct and possibly conflicting needs. In most models of learning, a monolithic agent tries to maximize one value that measures how well it balances its needs. However, this task is difficult when the world is changing and needs are many. Here, we considered an agent as a collection of modules, each dedicated to a particular need and competing for control of action. Compared to the standard monolithic approach, modular agents were much better at maintaining homeostasis of a set of internal variables in simulated environments, both static and changing. These results suggest that having “multiple selves” may represent an evolved solution to the universal problem of balancing multiple needs in changing environments.

Abstract

Satisfying a variety of conflicting needs in a changing environment is a fundamental challenge for any adaptive agent. Here, we show that designing an agent in a modular fashion as a collection of subagents, each dedicated to a separate need, powerfully enhanced the agent’s capacity to satisfy its overall needs. We used the formalism of deep reinforcement learning to investigate a biologically relevant multiobjective task: continually maintaining homeostasis of a set of physiologic variables. We then conducted simulations in a variety of environments and compared how modular agents performed relative to standard monolithic agents (i.e., agents that aimed to satisfy all needs in an integrated manner using a single aggregate measure of success). Simulations revealed that modular agents a) exhibited a form of exploration that was intrinsic and emergent rather than extrinsically imposed; b) were robust to changes in nonstationary environments, and c) scaled gracefully in their ability to maintain homeostasis as the number of conflicting objectives increased. Supporting analysis suggested that the robustness to changing environments and increasing numbers of needs were due to intrinsic exploration and efficiency of representation afforded by the modular architecture. These results suggest that the normative principles by which agents have adapted to complex changing environments may also explain why humans have long been described as consisting of “multiple selves.”

PIKA LABS site: https://www.pika.art/demo

A protein secreted by seemingly dormant cells in skin moles causes hair to grow again. That’s a big—and potentially useful—surprise.

Looking for a self hosting solution or some free alternatives for either chatgpt 4 or ideally sudowrite.

The latter does most of what I want it to do, but the free trial is super limiting and the paid tier is asking too much for not much of a word bump.

The former is only available with a sub but it requires way too much fiddling considering I'd be paying 20$ a month to do what I want but is an option. Sadly, I need gpt4 if I go that route and can't cruise on 3.5 for free.

I'm writing a novel and while I don't care about AI doing the actual writing for me, I do want something to help me organize my ideas or even brainstorm. gpt 3.5 just doesn't have the token bandwidth to do that. Sudowrite does an excellent job with it, but the pricing is stupid at 10$ for 30k words. I went through the 4k free trial just trying to figure out how it works.

I know there's a slew of self hosting chatbots but I haven't seen anyone use them for writing and searching huggingface is a pita.

Google bard could be an option but haven't found a way to jailbreak it and Claude is not available in my country.

Any ideas?

Bing (multimodal) image input is free!

Apple designer and Humane cofounder Imran Chaudhri envisions a future where AI enables using a phone without a screen

https://www.businessinsider.com/ai-chatbot-ceo-laid-off-staff-human-support-2023-7?IR=T

Abstract:

Large language models (LLMs) have demonstrated impressive results in developing generalist planning agents for diverse tasks. However, grounding these plans in expansive, multi-floor, and multi-room environments presents a significant challenge for robotics. We introduce SayPlan, a scalable approach to LLM-based, large-scale task planning for robotics using 3D scene graph (3DSG) representations. To ensure the scalability of our approach, we: (1) exploit the hierarchical nature of 3DSGs to allow LLMs to conduct a semantic search for task-relevant subgraphs from a smaller, collapsed representation of the full graph; (2) reduce the planning horizon for the LLM by integrating a classical path planner and (3) introduce an iterative replanning pipeline that refines the initial plan using feedback from a scene graph simulator, correcting infeasible actions and avoiding planning failures. We evaluate our approach on two large-scale environments spanning up to 3 floors, 36 rooms and 140 objects, and show that our approach is capable of grounding large-scale, long-horizon task plans from abstract, and natural language instruction for a mobile manipulator robot to execute. We provide real robot video demonstrations and code on our project page sayplan.github.io.

paper: https://arxiv.org/pdf/2307.06135.pdf

Video: https://cdn-uploads.huggingface.co/production/uploads/6258561f4d4291e8e63d8ae6/d_U_pzeCoJ2dTcBWz6n0r.mp4

Bard is available in new places and languages

• What: Bard is now available in over 40 new languages including Arabic, Chinese (Simplified/Traditional), German, Hindi, Spanish, and more. We have also expanded access to more places, including all 27 countries in the European Union (EU) and Brazil.

• Why: Bard is global and is intended to help you explore possibilities. Our English, Japanese, and Korean support helped us learn how to launch languages responsibly, enabling us to now support the majority of language coverage on the internet.

Google Lens in Bard

• What: You can upload images alongside text in your conversations with Bard, allowing you to boost your imagination and creativity in completely new ways. To make this happen, we’re bringing the power of Google Lens into Bard, starting with English.

• Why: Images are a fundamental part of how we put our imaginations to work, so we’ve added Google Lens to Bard. Whether you want more information about an image or need inspiration for a funny caption, you now have even more ways to explore and create with Bard.

Bard can read responses out loud

• What: We’re adding text-to-speech capabilities to Bard in over 40 languages, including Hindi, Spanish, and US English.

• Why: Sometimes hearing something aloud helps you bring an idea to life in new ways beyond reading it. Listen to responses and see what it helps you imagine and create!

Pinned & Recent Threads

• What: You can now pick up where you left off with your past Bard conversations and organize them according to your needs. We’ve added the ability to pin conversations, rename them, and have multiple conversations going at once.

• Why: The best ideas take time, sometimes multiple hours or days to create. Keep your threads and pin your most critical threads to keep your creative process flowing.

Share your Bard conversations with others

• What: We’ve made it easier to share part or all of your Bard chat with others. Shareable links make seeing your chat and any sources just a click away so others can seamlessly view what you created with Bard.

• Why: It’s hard to hold back a new idea sometimes. We wanted to make it easier for you to share your creations to inspire others, unlock your creativity, and show your collaboration process.

Modify Bard’s responses

• What: We’re introducing 5 new options to help you modify Bard’s responses. Just tap to make the response simpler, longer, shorter, more professional, or more casual.

• Why: When a response is close enough but needs a tweak, we’re making it easier to get you closer to your desired creation.

Export Python code to Replit

• What: We’re continuing to expand Bard’s export capabilities for code. You can now export Python code to Replit, in addition to Google Colab.

• Why: Streamline your workflow and continue your programming tasks by moving Bard interactions into Replit.