A beginners guide to AI: a risk/benefit analysis
In a world increasingly driven by technology, the emergence of Artificial Intelligence (AI) has undeniably sparked a shift in how we traditionally think about the intersection of technology and our lives.
Once the subject of science fiction, AI is now a tangible force with the potential to reshape the way we live, work, and interact. Where other recent headline-grabbing technological advancements, such as NFTs and the Metaverse have struggled to establish use cases with broad appeal, the possibilities of AI and how it may improve our lives are far more apparent.
The revolutionary potential of AI is hard to deny. From health care and data analysis to retail and customer service, there are very few areas of our lives where it is impossible, at least at a surface level, to make a case for the integration of AI. Jakob Nielsen of Nielsen Norman Group called it the first new UI paradigm in 60 years, and even after Facebook's very public rebranding to Meta, the company has all but ditched its focus on the Metaverse in favour of pursuing AI applications.
But as with all new technologies, traversing the discussion around AI is made difficult with unrealistic promises and misinformation. All of which can lead to disastrous results, as one US law firm discovered to their dismay.
To get the most out of AI, knowing its capabilities, limitations and potential downsides is crucial. Here we aim to illuminate the AI jungle before you and help you think critically about how this revolutionary new tech may find a place in your business.
The brief history of AI
While the sudden influx of media and industry attention may have you believing otherwise, AI is not a new technology.
Since the 1950s, computer scientists have been developing machine-learning algorithms. MIT's Joseph Weizenbaum developed Eliza, a natural language processing algorithm that is a spiritual precursor to today's ChatGPT as far back as 1964.
Despite the obvious promise of such developments, technological limitations meant it would be many years before AI would see large-scale adoption in any industry.
Fast-forward a couple of decades, and suddenly AI is everywhere, just perhaps not in a form you would typically consider AI.
Zoom blurring your background on video calls without a green screen; your iPhone responding to "Hey Siri"; Google Maps showing you the fastest route to your destination while factoring in traffic. All the applications above, and many more you interact with on a daily basis, rely on AI algorithms and have done so for years.
So if AI has been with us for years, why all the attention now? The most obvious answer is OpenAI. The release of the research group's AI-powered chatbot ChatGPT was a critical moment in the development of artificial intelligence. Finally, AI was no longer hidden away in the background of your favourite apps; users could now directly interact with the technology using plain language text prompts. In the words of Bill Gates - "The Age of AI has begun".
How does AI work?
Many of the fears around the technology (not to mention many of the outlandish promises made by AI promoters with something to sell) stem from a lack of understanding of what AI is and isn't capable of doing.
To understand AI's potential and limitations, we need to take a moment to discuss how AI works.
The term AI, as it currently stands, is a misnomer. Artificial Intelligence is not, in any way, intelligent; instead, it appears to be intelligent. And that fact is significant.
At its heart, most AI relies on pattern recognition. Developers train their systems on vast amounts of real-world data. AI algorithms then analyse this data to identify patterns, relationships and rules between the different elements of the data set. The patterns and relationships discovered by this process then form the basis of the AI's response to relevant inputs.
All of this may sound no different to the way human beings learn, but the difference is that AI does not understand meaning or context. It just understands patterns.
When you ask ChatGPT a question, it isn't answering it by thinking of an answer and then communicating it back to you. Instead, it responds with a sequence of words that its training data suggests is the most appropriate given the input.
This aspect of AI functionality is why programs like ChatGPT frequently suffer from 'hallucinations' where they state 'facts' that are demonstrably false or struggle with complex maths problems. Again, the AI isn't answering your question, as much as it is finding what its training data tells it is likely to be an appropriate string of words responding to your query.
Ultimately, at this stage at least, AI does not understand the meaning of its output. Whether it's words or images, AIs like ChatGPT manipulate and respond to symbols of meaning, not the meaning itself.
Does this mean AI doesn't have benefits? Not at all, just that you need to view AI in the same context as any other tool - it's not human; it has non-human limitations and capabilities that lend itself to particular applications. AI is not the go-to answer for everything, but it can be very useful when applied correctly.
Real-world business applications for AI
With so many possible applications of this technology, it can be overwhelming trying to understand how AI can benefit your business. While far from a comprehensive list, here are some real-world examples of AI in a business context that are achievable right now.
Chatbots
Given the dominance of ChatGPT in the conversation around AI, this is the most obvious application. Utilising AI chatbots can reduce client wait times, increase customer satisfaction and enhance the scalability of your customer service team. You can even limit bot responses to specific areas so that a human operator can take over in the event of more complex or nuanced questions.
Chatbots can also act as an approachable avenue to broach sensitive subjects the user may not be comfortable speaking to a person about. An example of this is Tundra's work with RMIT in developing Umibot, an AI-driven chatbot (using Amazon Lex) designed to help people subjected to image-based abuse.
Website data analysis
The release of Google Analytics 4 (GA4) has provided a promising look at the capabilities of AI from a data analysis standpoint. In GA4, algorithms automatically monitor your website data and alert you to data outliers that may represent opportunities or issues to address.
Content support
While many fear the potential impact of AI on creative jobs, content professionals can use AI to support their day-to-day activities and speed up workflows. Programs like ChatGPT can help with ideation by allowing you to quickly develop article structures or provide starting points to inspire your writing. ChatGPT also allows you to produce content at scale. Yes this content will need human input for fact checking and finessing copy, but AI can considerably boost writing efficiency.
To learn more about this specific side of AI please read our AI Chatbots for Content Professionals guide.
Smart supply chains
The pattern and trend recognition capabilities of AIs can be put to powerful use in supply chain management. Machine learning algorithms can analyse trends in supply and demand to provide forecasts for what is needed when — particularly useful for just-in-time supply chain strategies. For example, United States Cold Storage uses AI to analyse driver and route data to schedule accurate delivery appointments that reduce the risk of costly delays for the business.
Personalised recommendations
Amazon has used machine learning for years to drive a personalisation engine that it claims is responsible for almost 35% of its sales. Spotify made a splash in the market in large part to the effectiveness of its AI-driven song recommendation system. Netflix uses AI to determine not only what shows it recommends to you but also the specific cover image designs to use in doing so. Again, AI's power to spot trends in data makes it perfect for this particular application.
Auto-captioning
If you want to make your offering available to as broad an audience as possible, accessibility is vital. AI already provides benefits in this space with voice control for users with motor control and related disabilities. Furthermore, voice-to-text AIs make supporting hearing impaired users more manageable by automatically generating captions for video and audio content. For website owners without the resources to manually enable ongoing accessibility tasks, AI can be a saviour.
Marketing optimisation
Artificial Intelligence can be a powerful tool for managing pay-per-click campaigns like Google Ads. AI can significantly scale up the ability of advertisers to design and test large volumes of ad creative. AI can analyse user behaviour to identify and optimise for specific audience segments. Moreover, analysis and optimisation can happen in real-time, which is all but impossible for a human to achieve.
Semantic search
Search functionality is an expected feature on many websites. From large information sites such as Wikipedia to small eCommerce websites, a positive search experience can lead to enhanced engagement and retention. AI-driven semantic search enhances the search experience for users by identifying the contextual meaning of a search query and responding appropriately instead of just displaying a list of items that contain words found in the query.
What are the downsides of AI?
Copyright concerns
Programs like ChatGPT, DALL-E and Midjourney can achieve their impressive feats of technology thanks to the enormous data sets they are trained on. Of course, this data has to come from somewhere, and more often than not, AI scrapes data from the output of creators who have not provided their consent for their work to be used that way. Copyright holders have already found AI image generation programs incorporating their protective watermarks into AI creations. This specific aspect of the technology has led members of the 2023 WGA writer's strike to disparage AIs as plagiarism machines.
Privacy and security risks
AI systems often rely on vast amounts of data to operate effectively. This raises concerns about the privacy and security of personal information. If not handled properly, sensitive data can be vulnerable to breaches, misuse, or unauthorised access.
Lack of transparency and accountability
Some AI algorithms, such as deep learning models, can be complex and challenging to interpret. This lack of transparency raises concerns about accountability and understanding how AI systems reach their decisions or predictions, especially in critical domains like healthcare or finance.
Ethical considerations
AI raises ethical dilemmas and challenges. For example, driverless cars may face difficult decisions in life-threatening situations. Determining who is responsible for the actions of AI systems and establishing ethical frameworks for their development and use is an ongoing challenge.
Over reliance and our human dependency
Over reliance on AI systems without appropriate human oversight and intervention can lead to complacency and diminished human skills. Dependence on AI can also pose risks when AI systems fail or make errors, potentially causing significant disruptions.
Unintended consequences
AI systems can exhibit unexpected behaviours or produce unintended outcomes. Poorly designed or inadequately tested AI systems may have unintended consequences that can be difficult to anticipate or mitigate.
Perhaps the most famous example of an AI operating in an unintended manner would be Microsoft's Tay - an AI driven chatbot which lasted one day on social media before becoming so overtly racist that Microsoft shut it down.
Addressing these downsides requires ongoing research, the development of strong ethical guidelines, regulatory frameworks, and responsible AI practices to ensure that AI technologies are developed and deployed in a manner that benefits society while minimising potential risks.
Summary
While AI holds immense promise and potential, it is vital to approach its development and deployment with caution and consideration. Addressing job displacement, biases, transparency, data privacy, ethics, and system malfunctions is crucial for responsible AI implementation. By acknowledging these concerns, fostering public discourse, and implementing robust regulations, we can strive for a future where AI benefits society, while mitigating its negative impacts.
If you want to explore AI's possibilities for your business, free of overblown promises and hype, speak to us today.
Let's work together
Get in touch