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

Windows Copilot leak suggests deeper assimilation with Windows 11 features

Key Windows 11 features may soon be customizable as Microsoft further integrates its Windows Copilot AI assistant into the operating system. This tidbit comes from tech news site Windows Latest , which claims to have discovered new .json (JavaScript Object Notation) files within recent preview builds of Windows 11. These files apparently hint at future upgrades for the desktop AI assistant. For example, a “TaskManagerService-ai-plugin.json” was found which is supposedly a “plugin for Task Manager integration”. If this ever comes out, it could give users the ability to “monitor or close running apps using” Copilot. In total, six are currently tested and they affect various aspects of Windows 11. Next, there is an “AccessbilityTools-ai-plugin.json” that gives Copilot a way to “control accessibility [tools]. This would make it "easier for those with [a] disability to navigate through the system.” Third is “ai-plugin-WindowsSettings.json” for controlling important Windows 11 set...
Post a Comment
Read more

Google Chrome releases security fix for this major flaw, so update now

Google says it has fixed a high-severity flaw in its Chrome browser which is currently being exploited by threat actors in the wild.  In a security advisory , the company described the flaw being abused and urged the users to apply the fix immediately.  "Google is aware that an exploit for CVE-2023-2033 exists in the wild," the advisory reads. Automatic updates The zero-day in question is a confusion weakness vulnerability in the Chrome V8 JavaScript engine, the company said. Usually, this type of flaw can be used to crash the browser, but in this case it can also be used to run arbitrary code on compromised endpoints.  The flaw was discovered by Clement Lecigne from the Google Threat Analysis Group (TAG). Usually, TAG works on finding flaws abused by nation-states, or state-sponsored threat actors. There is no word on who the threat actors abusing this flaw are, though. Read more > Patch Google Chrome now to fix this emergency security flaw > Emergency...
Post a Comment
Read more

Samsung's ViewFinity S9 may be the monitor creatives have been searching for

Originally revealed during CES 2023 , Samsung has finally launched its ViewFinity S9 5K monitor after nine long months of waiting.  According to the announcement, the ViewFinity S9 is the company’s first-ever 5K resolution (5,120 x 2880 pixels) IPS display aimed primarily at creatives. IPS stands for in-plane switching , a form of LED tech offering some of the best color output and viewing angles on the market. This quality is highlighted by the fact that the 27-inch screen supports 99 percent of the DCI-P3 color gamut plus delivers 600 nits of brightness.  Altogether, these deliver great picture quality made vibrant by saturated colors and dark shadows. The cherry on top for the ViewFinity S9 is a Matte Display coating to “drastically [reduce] light reflections.”  As a direct rival to the Apple Studio Display , the monitor is an alternative for creative professionals looking for options. It appears Samsung has done its homework as the ViewFinity S9 addresses some of...
Post a Comment
Read more