Demystifying ‘Sensory’ AI and Skynet
As part of The IN Group’s Executive Exchange series, they recently hosted a roundtable to discuss all things AI with technology experts, which included Lead Associate Partner, Gordon McMullan. Read Gordon’s full event wrap-up and explore his key insights below.
Recently I had the pleasure of hosting an event with some of the leading ‘C-Suite’ and consulting intellects around a topic that has stimulated conversation of a level I have never seen in nearly three decades in the world of technology, namely Artificial Intelligence, specifically Generative AI.
I can think of no other topic of technological advancement that has generated such emotive led responses. From sheer excitement regarding the potential presented (and where I personally closely align and have experienced in the world of care pathways, transportation, retail) to absolute dread as to the impeding doomsday (which I understand and feel a deep responsibility to speak to).
In a world where it is increasingly difficult to decipher truth from ‘misinformation’, here’s a view to address fact from Skynet fiction. To unpick, I’ll address AI by its three core and recognised capabilities:
First is ‘Narrow AI’ (or Weak or Realised AI) – this is the only AI that exists commercially today. It is based on the premise of human dependency, i.e. the AI is based on the training input by us mere mortals and strictly functions within these parameters, unable to perform outside of its trained task, aka reactive machine –Skynet will have to wait.
The next evolution is ‘Artificial General Intelligence’ (AGI) or Strong AI. It remains largely a theoretical concept where AGI will develop skills unprompted through human training. Some progress along the lines of Generative AI (Gen AI) is being made, termed ‘Theory of Mind’ AI. This is where machine learning progresses to correctly interpret and respond to the emotive state of responses, i.e. speech (tone, speed etc.), to answer personally to the feelings perceived from users, the next level of Gen AI. Gen AI for briefness is content generating AI based on speech, text (language), imagery (sight) and audio (listening) – hence why I term Gen AI ‘Sensory AI’.
Lastly is ‘Super AI’, the self-aware AI that can be considered to have cognitive abilities that surpass humans, with its own motives, need and beliefs, aka Skynet. This is not the AI of today but seems to make-up a lot of the discussions in social and tv media where AI is commonly presented as one of the most dangerous developments in our, if not all time. Afterall, even with Narrow AI, machine learning processes and interprets volumes of data and parameters way beyond any single or collective capability of the most highly skilled users.
The prolific adoption of AI applications today, such as ChatGPT (Generative Pre-Trained Transformer) whilst offering huge productive benefits, does represent a real-world risk from the use of purely unknown and unqualified data applied in many settings, industries and therefore AI’s direct influence on the decisions made at local, national and even global levels.
So, at the heart of the debate is how can we monitor and attest the output from AI virtual assistants? One clear direction I advise clients for enterprise adoption is to use the foundation models proffered by AI as accelerators to train AI on an enterprise’s known data sets to address IP, sensitive data, bias, hate, and misinformation guiding the decisions at board level. Two other factors to consider in parallel, is how to continually monitor data being used ongoing and the forensic analysis required to explain variances and new bias that may be introduced. It is vital that any AI implementation is aligned with the ethical standards and the values of the implementer and there is a transparency on the way that the AI is trained and its decision-making process. Not doing so means we will only reap the calamity of risk realised too late.
If you would like to understand more on the topic of AI, we will be hosting several AI events throughout 2024, considerations for effective AI integration and adoption; core AI use cases to drive efficiency and revenue; and addressing AI risks in heavily regulated industries. Follow us on LinkedIn to be the first to know.
If you would like to talk more about digital solutions, change transformation and how you can keep moving forward, contact the Definia team today.
