On June 16th, we hosted an "Ask Me Anything" style webinar where interested applicants had the opportunity to have their questions answered live by a panel of Starburst and NASA officials. The transcript of this session as well as the full-length video recording can be found below. We also held another another "Ask Me Anything" webinar on June 23rd at 1PM PDT (4PM EDT) for which the transcript and video recoding can be found here.
Q: It seems that there are multiple interpretations of the physics-based transfer of learning topic, is the interest in the non-transfer learning AI only in physics-based AI, or is the interest mainly in transfer learning not more general AI?
A: We are looking for something a little bit different than just application machine learning. You know we have met a number of our scientists who are now using Tensorflow in their work for example. That's something we would not consider to be novel. We are looking for novel applications. Transfer learning is one in particular which I don't think we have a lot of experience with. So it doesn't have to just be that, but if you feel it's something novel it's better to err on the side of submitting rather than not.
Q: Is the interest in earth observation and disaster management systems primarily in satellite-based systems or is aerial imagery, collection, and processing of interest as well?
A: We have an aircraft program in earth science as well, we'd be happy to include that. It's not a satellite-only program.
Q: Is the main interest in physics-based modeling of the phenomena being detected/observed? Or would there also be interest in AI that understands the physics of the collection process and collection platform trajectory as that relates to accurate collection in modeling of the environment?
A:Yes. We would be interested in the latter.
Q: For Earth observation applications, is this constrained to onboard data processing only, or does this include cloud-based AI software applications that run on collections of remote sensing imagery?
A: The focus area will include 2 components: on the ground and then on board. It is assumed that the novel sensor’s data analysis is trained on the ground using historical data. Then transfer learning is performed using the physical characteristics of the novel sensor. The transfer learning may happen on the ground or autonomously onboard, the model is assumed to then run autonomously on board. If needed, an appropriate cloud-based training can be included.
Q: As it relates to applications specifically in the mass spectrometry topic, are there certain applications that are of interest?
A: We don't have any limitations on the types of technology for the mass spectrometer. Certainly, if you're going to use AI to, say, technology to mind mass spectrometry data that's also very much acceptable.
Q: In terms of technology readiness level, is there a certain TRL that we should be aiming for? Should it be on the lower or higher side? Or can it be anywhere on the spectrum of TRL?
A: It can be across the spectrum.
Q: For the AI topic area, the overall question is how broadly interpreted can these topic areas be? Are enabling technologies applicable as well as processing units or anything like that is specifically enabling onboard AI?
A: Processing Units, well, it has to be something novel. It cannot be something we are buying off the shelf and applying to a problem. It has to really be, I hate to use the term game-changer. I don't want to be too restrictive but if it's something new, something that we haven't seen before and it addresses our problems, I would go ahead and submit it. The nice thing about this is that it's only 5 pages. If you're not sure whether or not to submit, it's probably better to submit. If we don't like it, don't worry we will filter it out.
Q: In the description of round 2 in the challenge overview it's mentioned, “or they may choose to submit a proposal to one of SMD’s 15 established technology programs". Can you please state specifically the names of these programs? Perhaps we can share a link directly to these programs is there a place where people can visit them?
Q: This is differentiating between round 2 and round 3, so the rules state that in round 2 the judging panel will select up to 10 finalists receiving best scores to receive prize awards of $20,000 each based on the same scoring methodology in round 1. It goes on to say round 2 prize winners will submit a white paper that describes their concept in more detail to be eligible for an additional $80k in prize funding. The judging criteria for round 1 will be repeated and the winners will be the top 10 selected for those proposals receiving at least 40 points from the scoring process. So, only 10 teams make it to round 2, does that mean they're automatically in the top 10 for round 3?
A: That means they are eligible for the 80k. All 10 are eligible for the 80k, and all 10 might get the 80k. If we do the judging on their second writeup and they get 40 points, they'll get the 80k.
Q: Are there strict limits to the 5-page threshold since diagrams etc.. would exceed yet help communicate complex solutions or would it be recommended to attach as an appendix?
A: We are not reading page 6. We have limited resources, we have a limited number of reviewers, we just want to hear an idea: it's an elevator pitch. If you can't give us the elevator pitch in 5 pages then you want to go to another solicitation perhaps.
Q: Are there any rules regarding IP assignment?
A: The companies keep the IP, according to the compete ACT which we are using to execute this. The IP will remain with the teams, however, your question here is: is your IP safe when you share it with us… and the answer is yes as civil servants we are bound without even requiring an NDA we are not allowed to disclose confidential information. You should mark the bathing particular that is confidential just so that there is no confusion. But generally speaking, we as civil servants are not allowed to and I believe all of our contractors we sign into that same agreement. So on both fronts, in the ownership afterward you maintain that and none of the evaluators are going to take your idea and go pursue it because we would be breaking the law and we don't do that….. I think we specified that we did not want proprietary information in these writeups and again it's an elevator pitch so if you have the secret sauce you don't have to tell us about what's inside the secret sauce. Just tell us you have it, tell us why you can prove that it works and later on we can deal with that but right now we just want to hear your ideas.
Q: Can we submit more than 1 proposal?
A: Don't recall anything in our rules that say you cannot and so if you want to submit 4 you know we can't guarantee which ones will win, or if any will win…. I don't have a problem with it myself.
Q: In some cases were bidding on various NASA SBIR and JAIC related opportunities, are there rules which prohibit the submissions of this challenge if concurrently submitted to SBIR or JAIC?
A: We are talking to all of those people so we will I would say send the proposal into us and we will walk the halls appropriately and determine the best thing to do. It's not going to be a risk of double funding or anything. Generally, they do not like to pay for the same thing twice. So SBIR they're part of this, so we will be checking very carefully for all the submissions that come in. We will be checking the database the author may have submitted. It has to be distinct. I mean we can't be funding the same thing with 2 programs like Steve said so we’ll make sure that doesn't happen.
Q: They just said they have received a prior SBIR phase one. I'm not sure if that was NASA or airforce, so say they received an Air Force SBIR, can they still apply here?
Q: Are there any rules regarding using NASA in their PR materials once they've “A” submitted or “B” received the round 1 prize money?
A: You can look upon NASA’s site for the guidelines about using the NASA insignia. I will say in the past lawyers have been very rabid about that and making sure that people did not use it inappropriately. I will say in this particular leadership is easing off on that. And so I will just leave it with that. We now have the worm back so there all sorts of new stuff going on out there. But if you want to stay strictly legal, make sure you look at the branding guidelines that NASA provides.
Q: They say they have a game-changing infrastructure solution that isn't perfect necessarily for any of the categories but it would bring major benefits to all of them. Should we apply under the closest category, in our case AI learning, and just make mention of the beneficial aspects that affect the other categories?
A: It doesn't matter, I mean whatever you think is best, we will figure out what to do with it.
Q: Are there restrictions in partnerships with foreign countries. Say it is a US-based entity but they might partner on some aspects with a foreign company, are there restrictions there?
A: I don't believe so, as long as you have the US component, there are certain laws that all US entities have to follow like you can't bring in North Korea on your team and expect that to fly. But it doesn't have to be a total all US team. We simply can only pay a US team, you know the core team has to be here in the US or here through a legal residency.
Q: What qualifies as an early-stage company or are there restrictions on the stage of the company?
A: Yeah, okay, the intent here is to support startups and early-stage companies, however, legally we cannot restrict this to anyone. And if Northrop Grumman were to submit a proposal, we would have to consider it, and if we liked the idea we would fund it. It is intended for early-stage startups but it is not restricted to them at all. And there is no penalty in the scoring if you're not an early-stage company.
Q: How then would a winner of this competition be able to submit for SBIR phase 1?
A: At the innovation opportunity conference, which is SBIR, you would learn quite a bit about how to propose for phase 1, I mean I think this is a nice on-ramp into SBIR. You would take your idea that you flushed out in this “phase 0” and then it could be great preparation for phase 1. They're not connected, there is no connection to them, but I think this would be a nice way to go from a prize competition into phase 1. I think it's a great approach ……. The thing on the comment I made on you don't want to pay for the same thing twice is you gotta recognize that you guys are actually progressing and maturing the technology. What we don't want is if the government has paid for something and already received it were not going to pay for that again. So we wouldn't pay for you to get into the same stage that you got in the SBIR. If you get awarded an SBIR phase 1 they are going to expect that you progress to a new level…. Sometimes in the tech programs we see recycled proposals. They submit one thing to one program and submit the same thing to another program just to kind of send it out there and that's fine, but as program managers, we have to make sure we don't fund the two activities that are the same thing at the same time. That would be not good for the taxpayers.
Q: Are there links available for samples of the types of data, and is it the earth observation data that you're looking at? Is it specifically that AI category? Basically, is it up to them to comb through the linked reports and the references to understand the types of data?
A: Earth Science data is available to the public here and the Kepler data can be found here.