Skip to main content

Defining fairness: How IBM is tackling AI governance

Enterprises are hesitant to adopt AI solutions due to the difficulty in balancing the cost of governance with the behaviours of large language models (LLM), such as hallucinations, data privacy violations, and the potential for the models to output harmful content.

One of the most difficult challenges facing the adoption of LLM is in specifying to the model what a harmful answer is, but IBM believes it can help improve the situation for firms everywhere.

Speaking at an event in Zurich, Elizabeth Daly, STSM, Research Manager, Interactive AI Group of IBM Research Europe, highlighted that the company is looking to develop AI that developers can trust, noting, “It's easy to measure and quantify clicks, it's not so easy to measure and quantify what is harmful content.”

Detect, Control, Audit

Generic governance policies are not enough to control LLMs, therefore IBM is looking to develop LLMS to use the law, corporate standards and the internal governance of each individual enterprise as a control mechanism - allowing governance to go beyond corporate standards and incorporate the individual ethics and social norms of the country, region or industry it is used in.

These documents can provide context to a LLM, and can be used to ‘reward’ an LLM for remaining relevant to its current task. This allows an innovative level of fine tuning in determining when AI is outputting harmful content that may violate the social norms of a region, and can even allow an AI to detect if it’s own outputs could be identified as harmful.

Moreover, IBM has been meticulous in developing its LLMs on data that is trustworthy, and detects, controls and audits for potential biases at each level, and has implemented detection mechanisms at each stage of the pipeline. This is in stark contrast to off-the-shelf foundation models which are typically trained on biassed data and even if this data is later removed, the biases can still resurface.

The proposed EU AI Act will link the governance of AI with the intentions of its users, and IBM states that usage is a fundamental part of how it will govern its model, as some users may use it’s AI for summarization tasks, and others may use it for classification tasks. Daly states that usage is therefore a “first class citizen” in IBM’s model of governance.

More from TechRadar Pro



Comments

Post a Comment

Popular posts from this blog

The latest Apple TV 4K test lets you watch four sports streams at once

Apple is trying something new with the latest beta version of tvOS 16.5: the option to watch up to four simultaneous streams at once. Right now it's limited to live sports streamed through the Apple TV app on the Apple TV 4K , specifically MLB Friday Night Baseball and the MLS Season Pass. A multi-view option was spotted in the tvOS software last month, but the code was hidden and not enabled. MacRumors reported that the feature would be enabled this weekend, and beta testers have since been able to use it. As yet multi-view hasn't been officially announced by Apple, but it's expected that tvOS 16.5 is going to be pushed out in its final form within the next month or so. WWDC 2023 is around the corner as well, when we should be hearing about the next major updates for Apple's various operating systems – including tvOS 17. How it works Over at 9to5Mac there's a hands-on demonstrating how the multi-view feature works, and it's pretty much as you would expe...
Post a Comment
Read more

Garmin's new radar-equipped tail light will keep you safe on your e-bike

Garmin's Varia bike radars are some of the most popular pieces of cycling tech around – and now the company has delivered its first rearview radar to have been specially designed for some of the best e-Bikes .   Garmin's Varia range mounts to the back of your bike and broadcasts a radar signal behind you, so you can get visual and audible alerts when something's overtaking you. Even better, the new Varia eRTL615 plugs directly into most e-bikes, with no battery required. Because the catchily-named Varia eRTL615 is also a tail light, it'll also make sure you're visible to other vehicles too, promising to emit a flashing or solid light that's visible from up to a mile away in daylight. To connect Garmin's new radar tail light to your e-bike, you'll need to pick the right Garmin adapter cable (which isn't included). You can buy power cables compatible with Bosch, Shimano, or USB-A terminals or connections, with more info on those available on Garmin...
Post a Comment
Read more

Revolution Software is using their own AI technology to remake Broken Sword

TechRadar Gaming is reporting live from Gamescom 2023 on the latest and greatest developments in gaming and hardware. Revolution Software announced at Gamescom 2023 that Broken Sword would be coming back, with Broken Sword - The Shadow of the Templars getting a full remake while a sixth title in the series is coming in the future too, under the title Broken Sword - Parzival’s Stone .  Speaking to TRG ahead of the announcement, Cecil talked about the studio’s plans for a Broken Sword remake and the sixth title in the series. Cecil is a larger-than-life character, who is able to talk about the studio’s plans with enthusiasm. It even carries a pocketful of stones to illustrate the plans for Parzival’s Stone , but he also talks about how Broken Sword - The Shadow of the Templars would be using AI to upscale.  Cecil wasn’t shy about the studio’s use of AI technology, but he gave a fairly robust explanation of why the game was using it. The AI technology will be used to upda...
Post a Comment
Read more