https://community.element14.com/technologies/ai-machine-learning/In computer science, AI is defined as the study of "intelligent agents." These devices absorb information about their environment and take actions. Join our AI Group for online learning, discussion of AI development platforms, and more.en-USTelligent Community 12https://community.element14.com/technologies/ai-machine-learning/f/forum/57284/using-antigravity-with-a-real-project-a-practical-workflow/238376Fri, 25 Sep 2026 16:10:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:1457dbb7-df7a-4bbb-883a-9866a31af373ankur608hi shabaz have been lately working with antigravity to fly through code generation for few hobby projects (a FIDO2 key built on esp32). Unfortunately, gravity still exists and so do human-in-loop to get a working MVP. Hoping the agentic route lets me automate everything right up until the final hours.. ;) https://github.com/ankur608/OpenKeyhttps://community.element14.com/technologies/ai-machine-learning/f/forum/57284/using-antigravity-with-a-real-project-a-practical-workflow/238316Tue, 22 Sep 2026 01:44:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:416eb55d-a731-473e-9304-3b9a8e581bd6shabazGot it working.. turned out that Antigravity really wanted to see particular responses even though they were for things I didn't support. Anyway, long story short, it can work, but I think it would need a lot of testing if using such a cut-down implementation as I did. (+) Connecting AI to the Physical World with MCP and a Pi Pico W - element14 Communityhttps://community.element14.com/technologies/ai-machine-learning/f/forum/57284/using-antigravity-with-a-real-project-a-practical-workflow/238309Mon, 21 Sep 2026 19:55:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:02f129e7-b38b-4cef-b9d7-dfe91b5c9dacshabazDecided to do a quick experiment using MCP with hardware. To recap, MCP is a protocol for allowing AI access to external hardware/software, acting as a sort of demarcation point between the LLM and the "dumb" outside world. The outside world portion has the "MCP server" function to perform, waiting for requests from the AI, and responding to them. The AI can request the capabilities at startup for instance, then control or retrieve information. The capabilities are described in plain language text that can be hard-coded in the capabilities response for example, so the AI then knows what things it can do or obtain. There are some online project examples, mostly using a proxy type device in-between the AI (MCP client) and the outside world, but probably the capabilities have changed, since it's possible to use plain HTTP at the MCP server. Therefore a Pi Pico W can directly be used, nothing in-between required to encapsulate or tunnel information. I don't have decent source code to publish because it's inconsistent* currently (I will do so if it is ever a sensible library people can actually use) currently I just hard-coded it to respond to requests to control a relay or to acquire a temperature (hard-coded to 23.5 deg C). Anyway, once it was on the network and I had its IP address: agy mcp add --type http pico_w_mcp http://192.168.1.159:8000/mcp (The text "pico_w_mcp" was hard-coded in the Pico W code). Then in Antigravity: (Obviously this is a pointless example!). But shows MCP is quite crude, but can work. * Regarding the inconsistency referred to earlier: The relay is a "tool" request in MCP world. The temperature sensor would be a "resource". There were occasions when the server was added (using the command line shown earlier) where Antigravity would not understand it had a "resource" it could use, until I explicitly told it to check that out. On those occasions, Antigravity tried all sorts of bizarre things.. trying to guess what endpoints and protocols to use to fetch temperature (I only know this because could see the poor Pico being flooded, observed via WireShark), whereas Antigravity has the info all along, but would not use it until I explicitly told it the server had provided resource detail and to use that. It could be a bug in the server code, I can't be sure without a load more testing unfortunately (and would need testing with other agents too). I have a couple of theories though, but pure speculation for now.https://community.element14.com/technologies/ai-machine-learning/f/forum/57284/using-antigravity-with-a-real-project-a-practical-workflow/238302Mon, 21 Sep 2026 13:46:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:0406bdbc-bd57-47a8-a2fe-67a1415b6152shabazI've encountered errors but nothing I found noteworthy so far; and it will to an extent be model dependent (I know you're aware, but for newcomers: that's a setting that the user controls, visible at the bottom-right of the screenshots. Recently I've been using the default high-performance 3.8 setting, just to see how much it could do for the $5). There's not a lot of vendor lock-in, so it is realistic to go window-shopping and try a few agents, so I may well join you and switch to a different one at some point. So far it's handled everything I've thrown at it (i.e. it has always fixed any issues I've found), but I've only used it with about half a dozen projects (a mix of embedded, mobile and desktop). Since everything is natural language based, all the .md files are reusable for any future switch. The project rules sounds super-useful, since Antigravity can be extremely trigger-happy, and make assumptions that might be technically valid but certainly do not align with your own strict preferences (and far better to have them in such a folder that could be reused in future for other projects too, rather than mixing them into the project requirements and repeated conversation with the agent). MCP* is effectively an API for interfacing to the outside world, so that humans don't need to be the eyes/ears/hands. Many MCP-based servers already exist and the agent can locate, install and run them and start interacting with hardware. Unlike with a lot of APIs, with MCP it is the description field that is most useful in the server responses, since the AI accepts natural language in the query for what capabilities the server provides (e.g. for an oscilloscope you could have a "give me a screenshot" description in the resource capability, and later when the server receives the resource request message, it would go off and do that, and similarly a "set volts per div" described action [known as a tool in MCP syntax] for the server to do could exist). It is a really simplistic translator giving the AI capability that to interface, it can change from normal English language, to an actual plain server request message, since devices in the outside world can't understand natural language as well as it does. I've been writing a few apps for Android, and noticed an MCP server automatically got installed, so that Antigravity (but could equally be any other agent since MCP is a standard) could access the mobile phone, without needing me involved. * I don't know who was responsible, but surely whoever picked the name, was having fun and must have known what they were suggesting when they reused that acronym : )https://community.element14.com/technologies/ai-machine-learning/f/forum/57031/ia-generador-neural-network-library-for-microcontroller/238297Mon, 21 Sep 2026 11:09:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:af11009e-6721-4dbd-af1e-f473e22d0000BigGInteresting and I can see this helping others, especially in the teaching world. It would help if you provide a bit of history and what's the background to this project. Based on the contact email, you are involved in education/teaching and robotics. Otherwise, a key issue will be to explain what happens to your data. Is it processed locally or is it on the cloud and then what. Where does your data go.https://community.element14.com/technologies/ai-machine-learning/f/forum/57284/using-antigravity-with-a-real-project-a-practical-workflow/238296Mon, 21 Sep 2026 10:51:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:21f5122a-8eb3-4ccb-8946-025f1629fb85BigGI used to be an advocate for Antigravity until it failed dismally on knowing how to implement truth tables. I subsequently switched over to Codex and although slower it certainly is more meticulous, and I'm finding it better at embedded stuff. It also solved the truth table issue when generating code for an ESP32-S3 to act as a bridge device to flash another esp32 mcu (a blog is on the cards for this), although I was surprised at how much trouble truth tables gave Codex as well. The cool thing was that Codex was able to read my ADALM-2000 oscilloscope data logs and solved the problem that way. After many months of using Gemini AI and Antigravity, this is how I found the logic works: You're a new employee (whose name is Antigravity) & it's your first day. You see some flowers on the left side of the reception desk as you walk in. The next day you arrive, you see the flowers on the right hand side. So you walk over and move them back to the left side. Next day, they are on the right, and you move them back to left... and so it continues on the next day... TBH, I found this behaviour rather annoying. It's as if it expects the first review of your code to be a template to which it sticks to. So you have to explicitly explain every change you make, even if it's not part of the problem, otherwise it will revert the code back to the original, as if it's housekeeping. Often it won't even tell you about these changes. Fortunately there is a way to resolve matters. To reduce this problem and to improve matters when developing code I discovered some commands or tools that helped (these or similar commands apply to other agents as well). One is the "/learn" command and the other is the SKILL.md file. They have slightly different purposes (as summarised by Gemini): SKILL.md (Capability & Execution): A deliberate, pre-authored bundle containing explicit step-by-step instructions, constraints, and optional CLI helper scripts. It is invoked on demand via progressive disclosure when a task matches its trigger description. /learn (Extraction & Memory): A meta slash-command executed after or during a conversation. It instructs Antigravity to analyze the current session transcript, extract patterns, conventions, or bug fixes, and synthesize them into durable rules or candidate skills. In practice, you typically use the "/learn" option as a generator or feeder for your skills files or rules pipeline. Then there is another derivative: Project Rules (.agent/rules/). This is where you define the boundaries of your project - so things like "DO NOT TOUCH THE FLOWERS, NO MATTER WHERE THEY ARE" kinda thing. These are persistent instructions (coding conventions, architecture constraints, forbidden patterns) that the agent must observe across every turn of a conversation. Finally, there is this new concept that's supposed to help all AI agents called a Model Context Protocol (MCP) server. This is like a URL which points to a set of rules. For example there is a MCP for Zephyr RTOS. You assign this inside Antigravity and then it is suppose to give it proper context to the version you want to apply etc. I don't know too much about this, but there is a lot of online chatter about this.https://community.element14.com/technologies/ai-machine-learning/f/forum/57284/using-antigravity-with-a-real-project-a-practical-workflowSun, 20 Sep 2026 23:30:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:5956a15b-2c84-479d-be47-0c264a02d843shabazTable of Contents Introduction Get Your PC Ready Launching Antigravity Project Requirements Building and Testing a Skeleton Project Implementing the Project Documentation Testing Summary Introduction Judging from some recent comments, I’m not the only one curious about how people actually use AI tools to get things done. In this blog post, I decided to walk through how one assistant, called Google Antigravity, can be used. I cannot claim this is the best way to do things; there many excellent approaches. This blog post puts some text around some screenshots and comments I placed in earlier discussion threads, and also, the specific project example (which happens to be a real project) I use in this blog is immaterial; the goal is to discuss a workflow/procedure, rather than the actual project. Also, (just in case it needs to be said) this blog post was not written by AI - all mistakes are my own. Get Your PC Ready It possibly helps to keep folders organized and with sensible human-understandable names. The AI assistant can peek around (if you give it permission) so you want to help it. Here are the contents of my C:\DEV folder; I create a folder inside the projects folder, for any new project I wish to work on. You will require a Google account, but that is free. Next, I opened a shell (a normal Linux shell, or, in my case, PowerShell on Windows) and typed a command to install Antigravity (scroll down to the section Antigravity CLI on the Antigravity webpage ; no need to download directly from the page, unless you want to work in a graphical environment; I cannot comment on that, since I have not tried that!). CLI is in many ways easier, because it encourages one to script everything (or get AI to write the scripts of course) to simplify the automation of building things, and the speeding up of testing them. For instance, let's say you're working on an Arduino project. You could click on the sideways arrow in the Arduino IDE, to build and launch your code. If you wish Antigravity to be able to do that, then it's a lot easier if you install Arduino CLI ; then, whenever the code needs to be built (or even run), you (or Antigravity) can just issue a command from the shell, such as: arduino-cli compile "my_project_name" Launching Antigravity If you’re using the CLI version like I am, then the general method is, to go ahead and create a project folder like you may normally do for any project, and prepare as you wish. For instance, if you want to use particular formats, example data, custom hardware interfaces or scripts or configs, you could place them in the project folder. Placing them in a sub-folder could be a good idea, named (say) info or debug_scripts or example_data or anything that makes sense. Personally, I like to write up the project requirements in a file (make it a .txt or .md file perhaps, although other file formats can work), and place that in the project folder too (call it requirements.txt or phase1_requirements.txt for instance, but the name doesn’t matter). You could also write any conventions you have in a separate file (for instance, how you like code formatted). From the shell, go into your folder, and then type what looks like a typo: agy You’ll be prompted whether you trust the folder, hit Enter to accept, and a colorful logo screen will appear (it will look similar but slightly different to the screenshot, since Google likes altering things). Near the bottom of the screen is the prompt waiting for instructions from the user. Project Requirements I didn’t have any requirements written up beforehand for this particular project, so I instructed the AI to write the requirements into a file, in this case called GEMINI.md (but the name doesn’t matter). I typed the following in the command prompt: Antigravity can frequently prompt the user to allow or reject suggestions/changes. The default is allow , so you can keep you hand hovered over the Enter button mostly. You could choose to read the detail, or mostly ignore it if you don’t care about implementation in some areas of the project, beyond what you’d already put down in the requirements. How much should one write in an initial prompt or requirements file to the agent? The jury is still out, people have different approaches, but in practice Antigravity can consume as much as you provide it. It's not inconceivable to write a dozen pages even for a basic project, if you feel the project needs the detail, but at the same time you might not want to be so prescriptive for certain parts of the project if they don't matter as much, or, on the other hand, you may prefer a more incremental approach, splitting each phase of a project into lots of steps. Also, please note I'm using the word 'requirements' loosely, it is up to you if you take that literally to mean a project requirements list, or if you actually go further and state some of the high-level design, or if you request Antigravity to assist with that. I'd suggest you should also actually state the goal and and any example uses for the project, so that there is more context. I don't have an answer to what is the "best" approach, and there probably isn't a best approach. Building and Testing a Skeleton Project Whenever Antigravity pauses for a response, nothing will break if one selects No , even if the agent seems to want to do more. For some projects, it can be desirable to first prove things can compile (I’m assuming this is a software project for now, but it doesn’t need to be), so at this point I requested the AI to quickly get to a state where I can compile the skeleton code, even if it does nothing. Antigravity went off and searched for the tools installed on the PC (it requests permission for anything outside of the project folder) and made any changes needed to be able to compile the code (it made changes to a Makefile in my case), and voila, I had a binary executable all ready to run (on a microcontroller for this example project). Implementing the Project I typed in the prompt for Antigravity to proceed with the implementation based on the requirements. A snapshot of what it did is below; at that point it was reading some files and doing Google searches: After a while (it can take minutes or tens of minutes depending on the complexity): Antigravity had “completed” the project, and written up some notes in the shell, for me to scroll through and read. At this point I could install the built software and run and test it, and type up any problems at the prompt. Note that Antigravity can be extremely trigger-happy, and implement a load of work before you’re even ready for it. It can be controlled by explicitly typing as part of the prompt, to provide information first for approval, before implementation. Another simple approach is to use another LLM : ) I sometimes use Chat GPT for drafting what I wish Antigravity to do, and that way I can have a discussion with no immediate action from Antigravity, comfortable in the knowledge that ChatGPT isn’t talking to it behind my back. You can get to a point where you're merely a carrier pigeon between the two AIs. Documentation Since I was unsure what Antigravity had done specifically in a few areas of interest to me, I requested it to write some notes: After some thinking: Here’s a snippet of the final documentation it wrote: Testing In order to test the project, it is possible to request Antigravity to write and run test-cases. For hardware projects, if you’ve exhausted stubbed-out hardware test cases, then unless there is a way to provide stimulus and retrieve responses between the hardware and the AI, you’ll have to be the hands and eyes on behalf of it. Therefore, it does make sense to spend time automating if possible. You could use test equipment with USB or network interfaces. The AI is perfectly happy grabbing screenshots from test equipment on its own accord, spot issues, and fix and retest. For this simple project, I used the AI to write Python software for a manually operated test harness; it was admittedly a waste of time, when the AI could have tested by itself, with a little more effort from me setting things up. Summary Using Antigravity is a smooth experience; setup is quick, and it works from the command-line, which I prefer. You can start off with a blank project folder, or place some helper files in there, and that and some initial requirements were all that were needed to start building an app. It is worth thinking about how a project will be tested, and if anything can be done to remove the slow human from that loop. It was free to use Antigravity, but you can very quickly run out of credits using the free account. I noticed Google currently has an offer, charging about $5 per month for three months, before the cost is raised to $20, for the bottom tier. I think it’s worth it. It would be great to hear about how others use either Antigravity or any other agents. Thanks for reading!GeminiantigravityChatGPThttps://community.element14.com/technologies/ai-machine-learning/f/forum/57031/ia-generador-neural-network-library-for-microcontroller/238213Thu, 17 Sep 2026 06:58:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:f4b33845-a7fa-4f82-b28b-42872ab4ddf4hazratbilal[quote userid="522582" url="~/products/raspberry-pi/f/forum/57031/ia-generador-neural-network-library-for-microcontroller"]thanks[/quote] This is an interesting project, especially with machine learning moving toward smaller devices and microcontrollers. Lightweight neural network libraries seem really useful for processing data locally without depending too much on powerful hardware or cloud services. I’d be interested to see how these models handle limited memory and processing power in real-time applications. I’ve also been exploring practical tech and gaming projects at astutebetaserverhttps://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238095Fri, 11 Sep 2026 10:15:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:3901eaaf-e8f4-4555-8ef6-f444f3b9b8a6BigGLess is more, so to speak. There's one thing that is guaranteed. It will ALWAYS respond to you, no matter what. No silent treatment. No ghosting etc. I find that I am starting to use key action words, no matter the task. "Delve/Dive into...", "Expand on this...", "Explore...", "Investigate this problem...", and of course "Proceed with that recommendation/solution/suggestion..."https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238086Fri, 11 Sep 2026 01:19:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:24640003-8155-41e6-ba17-cf879cdc288fDABYes, I have been playing with it. Nothing earth shattering, but it will be some time before I can trust anything generated by AI until it is verified and validated. Since no one seems to be interested in verified results, I foresee a lot of bad decisions being made on junk information.https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238085Thu, 10 Sep 2026 20:24:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:4468b78a-8b4f-485f-b32f-eddead2a84acshabazI've been trying a few things just to see what's possible. got it to do this in one evening.. (a little game that my niece and nephew wanted, their mum wanted no ads or timers like the commercial games). I can't share it since the game is likely copyrighted (I based it on an existing game with cars, just replaced with boats : ) and also contains purchased childrens songs they like, playing in the background, hence I'd be violating some rights if it were distributed. Theyl zip through the puzzles (until they start getting harder), so it is hopefully a fun challenge.https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238084Thu, 10 Sep 2026 18:59:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:ccc160bc-a16a-4ef6-996b-e5e8a9999862Jan Cumps[quote userid="123345" url="~/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238083"]11) Write extremely long prompts if necessary, [/quote] I use a different technique (at this moment : ) . Have a conversation in short sentences - one-liners. Precise when needed, fuzzy if I don't have an idea myself yet. I like that incremental approach. (13). If you get an error, just paste in AI and ask what the likely causes (and maybe solutions) are. It's often right. Even when its own code throws a compile- or run-time error, pasting that error results in better/working code.https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238083Thu, 10 Sep 2026 17:49:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:c4ec40a0-cf68-4b1f-bdd5-aa1fdbcea3bfshabazI'm learning as I go, still not sure how optimal I'm being, but I noted some (seemingly/hopefully) plausible suggestions somewhere in an element14 discussion comment but can't remember where. Anyway, long story short, it was along these lines (and this is mostly just personal perspective, others will have better ideas I'm sure): (1) There is a dichotomy, on the one hand you want to provide decent explanation of your goal and objectives, but on the other hand you also want to split up things so the AI is not overwhelmed with unnecessary detail in parts, so sometimes more than one AI session could be helpful (2) If you're coding, then really split up your files (or tell AI how to architect) because otherwise you will find it hard to delve into the results, and AI may make more mistakes later (3) Try to speed up the compile/build and the testing, because you will be the slow part of the loop, not the AI. Heavily prefer command-line for all this (get the AI to help make it all command-line driven if you don't have prepared makefiles etc). (4) For hardware, consider describing what you want in text, not diagrams or schematics, and don't ask for schematics, but ask for descriptions. I appreciate this might become old-fashioned very quicky, since AI capabilities are improving rapidly (5) Ask for tools/libraries/architecture/languages advice from the AI, don't assume that the tools you were planning to use were the best idea. If it suggests a different language or library for instance, then I will strongly consider it for coding even if it means having to learn it slightly first (most languages have the same concepts so this doesn't take long) (6) Try to make the AI create tests or stub out things, to reduce the amount of effort when it comes to manual testing. Sometimes the AI will automatically run tests if it can (e.g. I believe Chat GPT will prepare and then run some Python code itself on it's own before providing you with the code) - but AI can go further of course, actually hooking in to real hardware so the user doesn't need to be the tester, the AI can do that and retrieve results from the real testbed, and iteratively improve things with no user interaction. (7) I find I get quite good results if I am very specific in describing behavior that is important to my needs, but deliberately give some free reign in other areas where I don't care as much. But see point 8 next. (8) Definitely inspect the code, not necessarily for bugs (the AI can identify bugs from described symptoms by itself usually) but to see how things were implemented, and push back for rewrites if it took a lazy approach. With hindsight, you'll see the areas where perhaps one should have been more specific in the request to the AI. For example, if you don't specify a desired data structure, it may use something entirely different. It is important to push back, because otherwise you may end up with code you cannot easily maintain. (9) Although AI is excellent at understanding when you're referring to something you have already referred to in a slightly different way in the past, recently I have been trying to be a lot more strict with my terminology. You can make up the terminology as long as it is consistent. For example, if there are a load of items in a system I probably would give them specific names or acronyms and their ports specific names too, and then refer to them consistently. (10) This is quite hard, but in general I try to not lock myself down too much, e.g, just because I cannot envisage a need or a scale for something, doesn't mean the code shouldn't be "ready" for it, thus I tell the AI things that might be needed in the future, but not needed now. I don't succeed always, because I get ideas and feature creep temptations later (not as much as an issue with hobby stuff of course). (11) Write extremely long prompts if necessary, For instance, it shouldn't be unusual to spend the day just writing the prompt, before even submitting that to the AI. Some prompts can be a dozen pages of one's thoughts/notes, even for hobby projects. (12) Ignore 1-11 as necessary : ) All is in flux, and you'll come up with way better guidelines for yourself over time - plus I am sure there are probably good guides elsewhere. I like "Code" magazine, I am learning a lot from that, well worth the subscription for the paper+digital copy.https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238080Wed, 09 Sep 2026 19:57:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:5391696f-fd24-4975-bb70-5e8812c91e89SensoredHacker0AI has succeeded. the userspace of AI is more of the fad. Im looking at AI as a means to get data. from that perspective, AI has been wildly successful. what if the user wants data about a proprietary thing? lol flag it!https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238079Wed, 09 Sep 2026 19:52:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:564583b2-08ce-4af9-9c37-89b587f01fb4SensoredHacker0all knowledge, and all knowledge worth knowing aren't even close to the same thing. I've been placed on the Tero project. Hey guys, why did we name it that? ( because someone in marketing thinks dinosaurs are cool, but wanted to avoid the PT spelling, for something like pterodactyl. ) We're really hoping that everyone else wasn't paying attention in Latin classes. still we could have looked into the etymology a little more for our heating system flagship product name. One way or another, we'll keep you warm! I dont think the promise of AI is in having to learn it. The objective from the start was to know you better than you know yourself. This little hiccup in the AI timeline where everyone on earth is feeding into the AI landscape is really useful training data, but isnt the real purposed of the thing init? Bonus! people will pay to participate! lol.https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/238057Mon, 07 Sep 2026 19:48:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:ec414e82-bd57-4486-988d-a321b6afc78dHKPhysicistI am glad this article has brought you some happy moment. I just concluded some of my and friends' ideas while I browsed youtube videos, news, etc. We must express different viewpoints.https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/237850Sat, 15 Aug 2026 00:18:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:3510bf93-80bf-495e-9bb0-000520be01cccolporteurIt is not my intention to monopolize the topic but I just found this quote and it is so how I feel. "Future shock is the shattering stress and disorientation that we induce in individuals by subjecting them to too much change in too short a time." quote by writer Alvin Toffler. It describes the psychological paralysis, anxiety, and disorientation people experience when the pace of technological and social change outstrips their ability to adapt. They is a picture of me next to this!https://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/237848Fri, 14 Aug 2026 23:56:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:3a22d351-0e8e-4b93-a7d1-c986d44d218dcolporteurK, I am more and more becoming a believer that if we do not learn AI we are at a disadvantage. Since joining ChatGPT Plus, I’ve had to confront many personal biases to maximize my effectiveness. I refused to upgrade my big-button cell phone until finally it prevented me from doing what I wanted to do. The same for my AI subscription. I had reached a point that no AI or free AI limited, me. I have 40 years of learning to apply when I use AI. I find it difficult to direct its actions toward my goals, as it asserts its own will. Microsoft unilaterally determines this approach, removing your choice. My quest at the moment is to learn how to talk to AI. If you could follow my problem-solving chats with ChatGPT, you would find the response. STOP: Tell me how to tell you how I want our interaction to occur. Since getting a subscription, I have discovered there are personal settings that i am using more often. Learning how to interact with AI appears to be a self-learning exercise. If you find someone willing to share their insight, it can make the interaction more fruitful. I wish I could still have my big button cell phone. That is not reasonable, but then again I liked wooden hockey sticks. You bring one of those into a dressing room today, no one will sit next to you. Bt the way. Great topic. I only wish it was with a bunch of folks around a table drinking coffee and trying to solve world hunger. P.S. I even used AI to proof this document so these are not my words. No that statement is not true. AI doesn't like my grammar or lack of spelling skills. My grade 10 teacher had the same AI issue.https://community.element14.com/technologies/ai-machine-learning/b/blog/posts/this-ai-chip-behaves-similarly-to-the-human-cerebellumFri, 14 Aug 2026 20:31:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:54fbb9b3-2c19-4569-bcfd-cac9e99a95aeCatwell(Image Credit: BrianPenny/ pixabay ) Northwestern University engineers developed an AI chip that behaves like a cerebellum. Rather than continuously analyzing, the chip reacts to unexpected events. During tests, the device detected arrhythmias within one-fifth of a heartbeat and with over 98% accuracy. The team says the chip could lead to low-power, always-on AI systems for wearable health monitors, self-driving vehicles, autonomous robots and cybersecurity systems. “In the world of brain-like computing, researchers typically try to mimic the cerebrum, which is often viewed as the brain’s ‘thought center,’” said Northwestern’s Mark C. Hersam, who co-led the study. “In our work, we developed a device that mimics the cerebellum, which controls reflex reactions seemingly without even thinking. The cerebellum is excellent at ignoring the expected and reserving its resources for reacting to the unexpected. That approach ultimately translates into lower energy consumption, and that is where we achieve orders of magnitude improvement.” The team is building on earlier work on improving AI hardware by combining memory and computation into a memtransistor. Modern computers consume a lot of power by shuttling data between separate memory and processing units. In a 2023 study , they demonstrated that two memtransistors performed AI classification requiring over 100 transistors while consuming 100 times less energy. Their AI chip goes even further. It replicates the cerebellum’s internal circuitry that detects unexpected events and makes quick decisions. Instead of analyzing routine information, it detects unexpected changes, including an irregular heartbeat in a wearable, a human stepping into a robot’s path, or potential cyber threats. “Today’s AI is remarkably good at recognizing patterns, but it often spends enormous amounts of computing power to continuously analyze streams of data — even when nothing has changed,” Hersam said. “Therefore, it burns energy on unnecessary analysis.” The cerebellum balances excitatory and inhibitory signals to determine if there’s unrecognized activity or unexpected events. Typically, those two signals offset each other. However, a sudden change affects that balance, prompting a response. With that in mind, the team replicated that behavior in a memtransistor with two operating modes. In one mode, the device’s response strengthens as stimulation continues. The other mode reacts strongly to an earlier signal before quickly fading. To switch between the two modes, the applied voltage must be reversed. The device’s behavior occurs due to its asymmetric transistor design. In this case, it uses an extremely thin layer of molybdenum disulfide and an electrode that partially overlaps it via a thin insulating layer. The device uses these features to distinguish routine activity from unexpected changes without constantly processing all the incoming signals. They used ECG recordings containing normal heart rhythms and arrhythmias to test the AI chip. It didn’t use up power by analyzing every heartbeat, ignoring the normal ones instead. The system also detected an abnormal heartbeat within milliseconds. The team wants to make the chip learn and adapt over time. For instance, repeated events eventually become familiar, causing the brain to no longer treat them as novel. “We have demonstrated one part of the cerebellum neural circuit, but there is more that we have not yet emulated,” Hersam said. “We intend to continue going down this path to mimic more and more of this complicated system.” Have a story tip? Message me here at element14.researchhmiartificial intelligenceon_campusmachine learningaibiomimicryuniversitybraininnovationhttps://community.element14.com/technologies/ai-machine-learning/f/forum/57161/this-is-how-i-compare-laplace-conjecture-with-ai-hype-conjecture/237797Mon, 10 Aug 2026 20:52:00 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:6ba7c5e4-b1b7-4dab-a370-7125f0658dbckmikemooI think "AI" is such an ambiguous and ill-defined concept that it can succeed and fail at the same time. Those segments that fail will be defined as hype and those that succeed will just become part of what we do - like the "decision systems" of the 1980's.