Some simple demos we did 3 years ago using our technology
https://youtu.be/NMHEynOLtKE https://youtu.be/7jIS5IUgVrw From: Stefan Reich via AGI <agi@agi.topicbox.com> Sent: Sunday, November 4, 2018 4:55 AM To: AGI <agi@agi.topicbox.com> Cc: a...@listbox.com Subject: Re: [agi] What kind of AGI research program would you propose? > To that end, lets say ou had ten million dollars (or some other large but finite sum), how would you invest that towards AGI? I would continue what I do - creating the ultimate user experience <https://www.meetup.com/Artificial-Intelligence-Meetup/events/255936823/> . That way, we can also help the AI learn by having it learn interactively together with users. 3 eyes (2 human eyes, a web cam and a screenshot module) see more than, uh... 2 eyes, right? =) Specifically, I want to build an interactive image recognition. User says: "Look at the round thing there", which activates the circle search module. And so on. Today's software interfaces are dearly lacking in AI. Cheers Stefan On Sun, 4 Nov 2018 at 07:50, Alan Grimes <alonz...@verizon.net <mailto:alonz...@verizon.net> > wrote: While I don't really have enough to eat right now, I can't stop thinking about AI. =\ Check out one of intel's few interesting products... Unfortunately it's server only, available only through OEMs and has some weirdness going on with it's PCI-E connector (it should only require an x8 slot but it requires a mechanically x16 slot...) Still, very intriguing hardware! https://www.intel.com/content/www/us/en/programmable/products/boards_and_kits/dev-kits/altera/acceleration-card-arria-10-gx.html https://www.micron.com/products/advanced-solutions/advanced-computing-solutions/hpc-single-board-accelerators AMD has done a really poor job of advertising the capabilities of their hardware in physics and/or AI simulation use. =\ Nvidia, of course, has a suite of products ranging from single boards up to massive rack systems. >From the software department, we don't have too too much to get excited about... https://arxiv.org/abs/1810.06936 https://labs.iuii.ua.es/3dperceptionlab/ It can be a useful excercise to consider what you would do if you had vastly more resources than you currently do. To that end, lets say ou had ten million dollars (or some other large but finite sum), how would you invest that towards AGI? What currently existing technologies would you use in your project? When in your project would you buy large scale hardware? What, in general terms, do you think you would get? What kind of people would you hire? How long would it take you to validate or invalidate your approach? For me, I'd invest in some high-end workstations equipped with a handful of high-end boards and some 16 core threadripper processors. I'd have some more hardware in the server room, enough to test and validate but not a whole-hog data center. I'd try to identify a handful of high aptitude individuals and give each the right to hire 2-3 underlings as both a "farm team" and to handle more mundane tasks. About half of each work week would be scouring journals, and other literature. Any young organization in hightech will need to acquire information FAST. The other half of the week will be getting technologies working in-house. As I mentioned previously, I see AI, neural interfacing and VR as being the pillars of the singularity. I'll give the team the freedom to work on whatever they'll be most effective at on any given day but the goal is to develop each to the point where synnergies can start happening and things can take off. I think VR can be a focus because games are Pretty Good, only a few more orders of magnitude on the physics front can really go a long way. I also think that putting a synthetic person together as soon as possible will be a critical step to identifying the missing pieces. On the AI front, I'd like to take some of the most promising systems such as Eliasmith's spaun (which was running at 1 second per hour) and see if I can optimize and scale it enough to operate on usable timescales. The fundamental problem is that there are a lot of ideas out there but it requires massive numbers of other things to be working in order to test them, being able to jump in and test sutff in an hour or two can be revolutionary. -- Please report bounces from this address to a...@numentics.com <mailto:a...@numentics.com> Powers are not rights. -- Stefan Reich BotCompany.de // Java-based operating systems <https://agi.topicbox.com/latest> Artificial General Intelligence List / AGI / see discussions <https://agi.topicbox.com/groups/agi> + participants <https://agi.topicbox.com/groups/agi/members> + delivery options <https://agi.topicbox.com/groups/agi/subscription> Permalink <https://agi.topicbox.com/groups/agi/T2aea40abd4ff90cb-M3fa34baa337dee6ce95b7c00> ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T2aea40abd4ff90cb-Mdf79af68500a6ff5f2aa8af5 Delivery options: https://agi.topicbox.com/groups/agi/subscription
