Notice

Scoping out the Grey Zone – Frequently Asked Questions

Updated 4 May 2023

Q: Following a successful bid, would DASA supply historical training data around these sub-threshold areas, or would data be synthetic?

A: We do not anticipate supplying any historical training data. We would permit the use of synthetic data. However, there would need to be an acknowledgement of the benefits and issues that come with the use of synthetic data in the proposal. A literature search may also illicit data literature for use.

Q: Is the intention of this competition to supplement/enrich existing Intelligence, Surveillance and Reconnaissance (ISR) tooling, or develop new standalone tooling specifically around these challenge areas?

A: It could be either. We would welcome submissions that utilise ISR related tools available to defence personnel that can be augmented to AI technology for sub-threshold. However, we are also looking for capabilities that defence don’t have, but that incorporate AI technology.

Q: Can you please provide a simple concrete example of sub-threshold ISR related to this competition?

A: There have been a lot of narratives that have emerged in both print media and social media. Although these may have been from a number of years ago there is still reference to them in current media. We want to understand how we can anticipate these narratives and how they might spread or be recycled. Furthermore, we are looking to understand how we detect misinformation and how do we combat these narratives.

Q: What is the minimum technology readiness level (TRL) DASA will fund for this competition?

A: TRL 3, proof of concept. We are not looking for paper based reviews. We would be looking for fundamental approaches that can be tested and evaluated at a basic level.

Q: Roughly how many projects do you expect fund in Phase 1?

A: There is £800k budget for this FY and we would anticipate proposals to be of £50k - £150k in value. However, that is not a fixed figure and we would look at each proposal on their own merit, cost wouldn’t be an inhibiter. We would likely fund 8 – 12 proposals within the budget for this FY, but this is not fixed.

Q: Does the scoring benefit proposals linked with academia? Does the scoring benefit small and medium size enterprises (SMEs) over large organisations?

A: No

Q: Does DASA ever fund projects with a timescale of over 6 months?

A: Yes, but this depends on the route you take to DASA funding.

Q: Is this call focused really on public domain information of the type found on the internet, or is it also interested in the electromagnetic environment?

A: We are interested in public domain information and we envisage that any proposal submitted would have that included. We are possibly interested in electromagnetic environment, but it should be clear in the proposal how it links to the particular challenge area.

Q: Is it an advantage or disadvantage to have received previous DASA funding in similar areas and use that IP towards this call?

A: If you have completed a DASA contract in a similar area and the idea fits with this call, you can submit a follow submission. However, you need to state clearly in the proposal how the idea will be advanced and how the previous experience would add to this proposal. Demonstrating value for money (VFM) is a key part of a DASA submission and this should be clearly evidenced in your proposal. We are not looking for a resubmission. If you can show clear evidence of further innovation and enhancement and the idea fits with the challenge are, we would be interested.

Furthermore, if you are involved with an existing Dstl project, then please make this clear in your proposal. While, under DEFCON705, the IP stays with you we need to ensure this is properly managed in the project. All proposal are assessed on their own merit, so previous or current involvement with Dstl will not provide any advantage.

Q: Is software delivery required?

A: Where a proposal includes software code, we would expect development of that. However, we wouldn’t expect that software code to be at a production level. We appreciate that it may not be possible to produce this within the constraints of the proposal value. Where code is generated we would expect it to be transferred to Dstl. While the intellectual property stays with supplier, Dstl and MOD would have user rights to the code in order to demonstrate the outputs to potential end users.

Q: Is it better to concentrate on one challenge or have proposals that apply to multiple challenges?

A: We welcome bids that link to multiple challenges. However, we do recognise that you may wish to focus on 1 challenge area and that wouldn’t be of detriment to you. Please specify, in the proposal, what challenge area your proposal is relevant to.

Q: Is there a required clearance level?

A: No, all work should be at UK Official. We need to know who is working on the proposal before awarding the contract and this can take time, particularly with non-UK nationals. Please build in to your project plan to allow for this process. Please refer to the DASA website for more information.

Q: Which end customers might this work end up with initially? (i.e. Army, Navy etc?)

A: Could be any and all of the above. There are a broad set of potential users across UK defence and wider HMG.

Q: Do compute and storage costs for related data have to be included within the total bid?

A: Yes. We would expect those costs to be included in your proposal, as there wouldn’t be extra funding available for this.

Q: Can non UK registered companies bid?

A: Yes. DASA has an international Innovation Partner, so please contact them.

Q: May you please advise how understanding intents and aims of an adversary differ to characterising narratives challenge problem?

A: There is a lot of noise in this arena. We would like to understand how we achieve early detection and then track the evolution of adversaries. In terms of understanding intents and aims, there are a lot of emerging narratives and some will appear more credible than others. The key difference is once we have detected the narrative we need to understand how we apply reason to the intent and aim. For example, is that to spread fog and increase counter time/debunk time? Or is the intent to target a particular audience? This would require to understand how we characterise the targeted audience.

Q: What proportion of projects to you expect to fund in each of the three challenge areas?

A: We will need to make a balanced judgement on how the overall aims of the project can be achieved. The focus won’t be on what proportion we have of funded proposal in each challenge area. Rather, we will be assessing each proposal on their own merit and this will inform our funding decisions.

Q: How does this theme call link in with the Defence AI Centre (DAIC)?

A: If there are submitted proposals that we deem fundable, but we do not have available funding for, we will be speaking to DAIC. We will work to ensure all good proposals are captured even if they are not submitted to what we deem the most appropriate stream.

Q: What is the appetite for getting Academia involved with these challenges (and what about clearance waivers/non-UK citizens)

A: We welcome bids from sole entities or universities and we also welcome bids that are collaborations with academia and industry. Each proposal is assessed on its own merit. More information on the DASA assessment process can be found here.

Q: Are there any deadlines for project completion? Following successful bids, what is the follow-on process? Are client resources on hand to work with, what does eventual delivery look like?

A: The projects funded through this competition would need to complete by March 2024.

There is potential for further funding next year, but this is not confirmed. There has been interested in this competition and potential outputs from across MOD and wider HMG.

Dstl colleagues will be on hand to support funded projects. Eventual delivery could take a number of forms and would be dependent on the work in your proposal. We don’t anticipate the delivery of a fully deployable product. We anticipate outputs to give an indication of what the ‘future could look lie’ if the technology was developed further. Much of this is dependent on your proposal.

Q: Is there a deadline for submissions?

A: Tuesday 6 June at 12 midday (BST).

Q: Will the Defence AI Centre (DAIC) own the delivery beyond initial phase?

A: Not necessarily, it may feed in to this Dstl project. However, the current AI programme is scheduled to end in May 2025, so we cannot provide confirmation at this stage.

Q: What’s the stands of DASA on generative text solutions like ChatGPT? Are solutions based on ChatGPT as a solution to the human-machine teaming goal acceptable?

A: There is some work currently being conducted on large language models and their impact on the work we do. This is ongoing and we are open to hearing suggestions on how these types of models could work in sub-threshold space. The key drawback is that licensing conditions of some large language models would prevent Dstl access to them. It is, therefore, important that you evidence in your proposal how you would address this challenge.

Q: Are there any specific technologies you are interested in applying artificial intelligence (AI)/machine learning (ML) to?

A: We are not looking to be prescriptive, as we want to see in the proposal how your idea addresses the challenges. We are interested in your proposal.

Q: You mentioned demonstrating this capability, would the output need to have a UI (User Interface) or could it be an API (Application Programming Interfrace) connected to established systems?

A: Where we have produced demos we have run a simple UI locally, so we have control over what we are doing and are not rely on external connections. If you can propose a good reason for why we would want to build in to an existing system, then we would be interested in seeing that.

Q: Is the goal to produce a solution with MVP (Minimum Viable Product) / software package, or more focused on generating a framework/conceptual solution?

A: To a certain extent, we want some system demonstrators. We are not interested in solely a think piece, we would like something tangible to demonstrate and work with. However, if a solution had low feasibility, but there was a clear development line we would be interested in that. Through the assessment process we will be looking at how innovative and conceptual a proposal is. While we have to be realistic about the anticipated outputs, we recognise some proposed ideas will be more polished that others.

Q: Is it possible for one organisation to submit multiple proposals against the different challenges?

A: Yes

Q: Can we submit our idea to DASA competitions which we already protected by patent filing?

A: Yes. Please include details of this in your proposal.

Q: Isn’t grey zone a pan government department activity? Couldn’t the end users also be Dept for Health, Treasury, Home Office, Foreign Office as well as MoD?

A: Yes. The MOD is funding this competition, so there needs to be a clear defence focus in any submitted proposal. Any use cases should link to relevant military activity. However, there are a number of examples and narratives the grey zone could apply to and there is scope for wider exploitation of any submitted proposal.

Q: Does this call relate to defence against ISR or support of ISR (or both)?

A: Looking at characterisation of ISR and how we conduct ISR against adversary activity

Q: Can non UK nationals living in the UK be involved in a collaboration with DASA?

A: Yes

Q: Will we have access to relevant users throughout the development of our ideas?

A: Yes. Those that are successful in receive funding with have access to Dstl personnel and indirect access to other end user groups.

Q: Are you looking for a fully managed service, software supported by human in the loop analysis?

A: We need to see clearly evidenced in your proposal how your proposed innovation (below TRL6) would be enhanced through receiving funding. Value for money is a key part of the assessment process and this should be clearly evidenced in your proposal.

Q: Will companies already working on the previous ISR competition be considered for this competition? Red Scientific was mentioned and some previous work.

A: Yes. All proposals will be asses on their own merit and companies we’ve have previous experience with won’t be given preference.

Q: For intend identification, is simulated multi-media detection data acceptable?

A: This answer depends on what is meant by simulated. In this answer, it will be interpreted as having created out own data artefacts rather than finding them in the ‘wild’. While we can find data artefacts in the wild, they likely won’t be reliable enough. In response, yes, simulated multi-media detection data is acceptable.

What are the existing ISR tooling that are used to look at the sub-threshold?

A: Can’t answer this today. Due to operational sensitivities it is not possible to release a list of the software toolsets employed in Defence and Security for ISR of the sub-threshold

Q: Is the main focus of this work on counter disinformation or are there other examples of sub-threshold ISR?

A: Disinformation is a very strong theme. However, there will be other projects and areas looking at other aspects.

Q: If we have existing partnerships with IaaS (Infrastructure as a Service) providers (E.G AWS) can we utilise them to cover storage and compute costs for the POCs (Proof of Concept)?

A: Yes. If you have an existing relationship with those suppliers, then you can incorporate them in to your proposal. Please provide clear details of the existing relationship in the proposal.

Q: Which TLB does the ‘AI Programme’ sit in?

A: The AI programme sits within Dstl which operates under Defence Science Technology (DST) and the MOD Chief Scientific Advisor. The closest TLB the AI Programme sits in is therefore Head Office and Corporate Services (HOCS).            

Q: What will you do with the software that is delivered to you? What is the plan for that?

A: The plan is to demonstrate the outputs of the research with end user organisations/units to highlight future capabilities. Depending on the level of maturity, it could lead to: Further research funding/phases to mature the technology as part of the AI programme; for more mature technical outputs, it could lead in the nearer-term to small-scale trials/evaluation with end users;

Q: How will the ‘valley of death’ be managed to transfer industry solutions to securing a TLB funded project via the MoD Annual Budgetary Cycle (ABC).

A: This will be dependent on the maturity level of the output and there are funding opportunities that sit outside of the AI programme that could help to mature technologies towards operational capability. Nearer-term demonstration and evaluation with end-user units/stakeholders will be an important first step along this pathway.

Q: How are you defining ‘relevant environments’ with respect to TRLs? Specific details would be helpful there.

A: A lower TRL (2-4) piece of development software isn’t something we would ordinarily look to install on end-user IT environments – such activities would likely need additional development following an early demonstration activity. If there are outputs at TRL5-6 which may be more suitable to progress to testing/validation with end-users, we may seek to use a representative environment or install on a suitable end-user IT environment. The latter would require further direct consultation between Dstl, the supplier and the end-user in question.

Q: Do we concentrate one type of data for example text or do we need to use multimedia data ie text and video and image for detection?

A: We would welcome submissions which address both single modalities and multiple modalities, recognising the time, cost and complexity increases with multi-modality data technology. In some instances, specialist suppliers may wish to focus on a single area of expertise. In others, obtaining multiple results and fusing them may be desired.

Q: Given the focus on AI in the technological solution, is there any guidance on how broadly AI is being defined here? e.g. statistical learning v deep learning

A: We see the range of possible solutions as being broader than just deep learning, so we would welcome proposals including deep learning, statistical, rule-based reasoning etc, we are not seeking to limit it to specific areas that may be categorised as AI.

Q: How do TRL levels align to alpha, beta phases? Are you looking at something meeting alpha stage at a maximum?

A: We see TRLs 1-4 as broadly corresponding to early alpha/, TRLs 5&6 to alpha, and 7 onwards to beta and beyond. We do note that there may be existing supplier solutions at beta or higher but that would need research/enhancements in order to meet the requirements of this themed call. For such proposals it would need to be clear what additional functionality/capabilities would be added if the proposal were to be funded.