Since we launched on Nov 27th 2019, we’ve been amazed by the number of requests, questions and comments that you’ve been sending our way! We want to thank you for the interest and please find below answers to some of the most common comments and questions. If there is anything else you’d like to know, please contact us at email@example.com and we will answer each specific email we receive.
Are you doing this manually? I would bet you have humans in the loop
No our system is fully automated (that’s why it’s a first) so bear with us as we are in beta :) Automation is possible because we use Reinforcement Learning (RL), an AI technology which is very useful for decision-making problems like Chess, Go, or PCB Routing. At InstaDeep, we deploy RL systems in the real world and work closely with hardware partners such as Nvidia (after a rigorous technical evaluation, Nvidia upgraded us to preferred partner, the highest status possible) and Intel (we’re part of Intel’s AI Builder Program). In our opinion, having no-humans-in-the-loop is critical to accelerate PCB development cycles, and that’s a key feature of DeepPCB™. To further illustrate that’s indeed a fully automated process, we are working on making available to you intermediary routing solutions: you will see by yourself our AI system progressively increasing the quality of its routing as it learns to solve your board.
This is an unsolved problem… It's NP-Hard, and you can’t solve it. It’s all AI hype and who are you anyway?
InstaDeep has built credibility in AI circles by innovating in Machine Learning and Reinforcement Learning (RL), specifically on how to find good solutions to NP-Hard problems with AI. For example, you can check our R2 paper https://arxiv.org/abs/1807.01672 about AI for combinatorial optimization which was accepted at the NeurIPS 2018 in the Deep RL workshop. More recently we’ve just published joint AI-research in RL with Google DeepMind that earned a top 2% global ranking at NeurIPS 2019 https://arxiv.org/abs/1905.12941. NeurIPS is the largest AI research conference in the world and is widely considered as one of the most important ones.
What do you know about hardware?
We have significant domain expertise in Hardware and Electronic Design Automation (EDA): several of our team members have worked many years in companies such as STMicroelectronics, NXP, Dialog, Sagem etc. We believe it’s an exciting time to be working on PCB design automation, and because we have both AI and Hardware expertise, we could be well positioned to do so.
What are you doing with my data? Are you going to sell it to competitors? If it’s free am I not the product?
InstaDeep will never share or sell any of your data to any outside entity. Our sole goal is to accelerate your PCB product development cycle. For now we have released a free beta version but we’re working on a commercial version to tackle larger and more complex boards. We anticipate keeping a free version to make it easier for hobbyists to design their boards.
Why does it take 24 hours? Results are so-so. Is something wrong?
DeepPCB™ is truly an innovative platform. To get results, it actually needs to spend time learning how to solve your board. The learning process is the fundamental difference between our system and others, and while 24 hours might appear long and inefficient, it won’t seem that way when we release upgraded versions of our platform that tackle more complex boards (we’re still in beta). As discussed before and to illustrate the automated learning taking place, we will release intermediary snapshots of the intermediary solutions the system generates during the 24hour period.
Auto routers have been there for ages and they’re useless anyway, so why are you doing this?
That’s the whole point, auto routers don’t get the job done properly because the problem is too difficult for the heuristics typically used by current software. This also explains why many boards are still designed manually. It makes sense that AI could improve things here and our product is a world-first: nobody was crazy enough to make publicly available a fully automated, pure AI product on the market until we did.
Routing at the end of the day is limited. The hardest part of PCB routing is Placement, do you plan to add Placement?
Yes we do. Our goal, once DeepPCB™ routing is out of beta, is to tackle placement. Routing is an important first step, but we don’t plan to stop there. In our livestream (https://www.youtube.com/watch?v=Ea5i-l8YKQo) we clearly mentioned that Placement is our key goal for the year 2020.
I have exhausted the initial 3 project limit trying to see how it would route, can you do something for me?
We hear you and have decided to increase the limit to 4 credits per week. We’ve also regenerated the credits for everyone already registered! We will periodically review credits based on demand to accommodate as many users as possible. If you would like more credits please contact us at firstname.lastname@example.org
This is not good enough, what’s the whole point?
Please keep in mind that we’re in early beta, and that our service is free. Even though we have for now strong limitations on the type of boards we currently support, our goal is to hear from you to iterate and improve the product quickly. Things might break from time to time and we’ll certainly make mistakes (we’re a small startup). What matters to us is to work hard and make progress on this exciting problem. We believe it’s the combination of hardware domain expertise and advanced AI know-how that yields strong results. If you have comments or would like us to improve specific features of DeepPCB™, please do not hesitate to reach out at email@example.com we’d love to hear from you!