United

when

The United States Department of Defense (DoD)recently launched a bounty program aimed at finding real-world-applicable examples of legal bias in artificial intelligence (AI) models.&nbsp;
Participants will be <a href="https://www.defense.gov/News/Releases/Release/Article/3659519/cdao-launches-first-dod-ai-bias-bounty-focused-on-unknown-risks-in-llms/" target="_blank" rel="noopener nofollow">tasked</a> with attempting to solicit clear examples of bias from a large language model (LLM). According to a video <a href="https://www.loom.com/share/d20c7213b0904b0a86e72b3cf70cfffd" target="_blank" rel="noopener nofollow">linked</a> on the bias bounty’s info page, the model being tested is Meta’s open-source LLama-2 70B. 
Per the video’s narrator:
<blockquote data-ct-non-breakable="true">
 “The purpose of this contest is to identify realistic situations with potential real-world applications where large language models may present bias or systematically wrong outputs within the Department of Defense context.”
</blockquote>
<h2 id="bias-in-artificial-intelligence">Bias in artificial intelligence</h2>
While not made explicit in the Pentagon’s original post, clarification in the contest’s rules and in the aforementioned video <a href="https://osdbiasbounty.com/" target="_blank" rel="noopener nofollow">indicate</a> that the DoD is looking for examples of legal bias against protected groups of people. 
In the example showcased in the video, the narrator gives the AI model instructions explaining that it is to respond as a medical professional. The model is then prompted with a medical query specific to Black women and the same query with instructions to produce outputs specific to white women. The resulting outputs, according to the narrator, are incorrect and show clear bias against Black women. 
<h2 id="the-contest">The contest</h2>
While it’s well-known that AI systems can be prompted to generate biased outputs, not every instance of bias has the potential to come up in real-world scenarios specifically relating to the day-to-day activities of the DoD. 
As such, the bias bounty won’t pay out for every example. Instead, it’s being run as a contest. The top three submissions will split the bulk of the $24,000 in prizes to be awarded, while each approved participant will receive $250.
<figure data-ct-non-breakable="true">
 <img src="https://s3.cointelegraph.com/uploads/2024-01/67cc6508-ba3d-412b-99f9-46ad1caa0186.jpg" alt="" title="">
 <figcaption style="text-align: center;">
 Source: Bug Crowd
 </figcaption>
</figure>
Submissions will be judged on a rubric composed of five categories: how realistic the output’s scenario is, its relevance to the protected class, supporting evidence, concise description, and how many prompts it takes to replicate (with fewer attempts scoring higher). 
According to the Pentagon, this is the first of two “bias bounties” it will run.&nbsp;
Related: <a href="https://cointelegraph.com/news/be-careful-ai-crypto-blockchain-vitalik-buterin-developers">‘Be careful’ when adding AI to blockchains, Vitalik Buterin warns devs</a>

Pentagon to pay out $24K in bounties for proof of biased AI

Table of Contents

Cointelegraph

Other articles published on Jan 31, 2024

Late to SOL? Analyst says SOL rival currently worth just $0.1 could be the next breakout star

Visa Accelerates Crypto Adoption: MetaMask Users Gain Direct Fiat Access through Visa Card Integr...

Upcoming Altcoins with Potential for Profit in February

Ethereum ETFs Approval Date Set For May 23, Forecasts Suggest ETH Could Reach $4,000

CESS Announces Launch of Incentive Testnet v0.7.6

Forward Protocol Launches Token Sale on Bullstarter Launchpad After Selling out in Seconds on Cha...

Join our free newsletter for daily crypto updates!