AI has made significant advancements in the past decade, now entwined in almost every aspect of our lives, from business to everyday activities. Despite current limitations, GenAI shows promise in empowering cognition and optimising workflow. Through human-centred design, augmentation offers a safer and more effective way to enhance human intelligence rather than replace it.
Amid AI’s tremendous hype, this hackneyed buzzword has weaved its way into just about every household from New York to Nairobi. With all this techno fixation one might be wondering why now, what’s all the fuss? Surely, not another crypto conundrum. You may even be perplexed to the verge of vexation by the frenzied obsession with ChatGPT and the clichéd mention of ‘Gen AI’ — report after report, course after course, boasting the mysterious might of an oversimplified acronym.
While grasping the complexity of deep learning with a weary eye for futuristic fad may be a rational rite of passage, the tectonic impact of this vast computing power is becoming crystal clear. Artificial neural networks are here to stay, and knowledge work is primed to benefit from Large Language Models (LLMs) as thought partners for empowered cognition.
The intricacy of AI
We’ve all pondered, in one way or another, about the immense potential of artificial intelligence. Yet how does this neoteric concept fit within the bounds of all the buzzwords flooded on our social feeds?
Recognising its intricacy requires a clearer picture of the integral parts: machine learning, deep learning and data science.
While AI involves machines or software that mimic humanlike cognition, machine learning encompasses statistical algorithms that learn from data to make generalised decisions with human intervention. Moving further into the algorithmic sphere, deep learning is a multi-layered artificial neural network that simulates the behaviour of a human brain and learns from large amounts of data based on representation learning — with or without human intervention.
Interlinking these components is the domain of data science, which applies statistics and scientific methodology to extract coherent insights from structured and unstructured data.
Intuiting this immersive technology, you may have also heard about the interplay between ‘narrow’, ‘general’ and ‘super’ artificial intelligence. The gist: as our lives become more entwined in deep learning, we can expect progress to move from a weak state of AI to a more advanced pervasiveness.
AI’s journey to prominence
Given huge machine learning leaps have been made in business and society, Artificial General Intelligence (AGI, i.e. machines surpassing human capability) remains, at best, a distant reality with numerous limitations affecting adoption. Think: inaccurate data analysis, algorithmic bias and adversarial attacks — all of which compound current scepticism about the viability of this nascent technology.
While current Artificial Narrow Intelligence (ANI) may not be able to fully replicate superintelligence in ways depicted by sci-fi epics such as Ex Machina and Her, there are promising signs of incremental progress relating to workflow optimisation.
Progress that McKinsey Global Institute estimates will culminate in generative AI challenging human level performance on business-related tasks at an intermediate level by the end of 2030, and competing with the top quartile of people performing such tasks before 2040.
Over the past decade ANI has reached users worldwide, most notably with natural language processing (NLP) as well as image and speech recognition systems. These systems are showing promising utility by enhancing cognitive dexterity through robust LLMs in the form of ChatGPT and Bard.
In the conversational LLM space, Siri, Alexa and Google Assistant have dominated the mobile app market while IBM’s watsonx and Microsoft’s Copilot (previously Bing) are gaining a firm grasp on the enterprise side.
With more and more data at scale, it now appears deep learning is permeating just about every aspect of our lives — from web search and language translation, to facial recognition, disease diagnosis, fraud detection, recommendation systems, and autonomous vehicles.
Generative AI abounds
Significant developments in transformer-based artificial neural networks have lowered barriers to entry for advancing deep learning applications. As LLM integrations become more affordable, organisations are beginning to make tremendous strides in optimising tasks and reengineering workflow through techniques such as fine-tuning and retrieval augmented generation (RAG).
Through this quicker workflow, humans are benefitting from various use cases of prompt-based AI, including: writing (suggestions or answers), reading (detecting complaints, or routing customer emails) and chatting (online orders via chat bots). Moreover, knowledge work, activities involving decision making and collaboration, stand to benefit the most from natural language processing.
With startups sprouting in just about any B2B/2C segment you can imagine, notable applications of GenAI include: spam filtering, online advertising, self-driving cars, healthcare, visual inspection, speech recognition, and reputation monitoring. All of these segments are being significantly disrupted, forcing existing value providers to rewrite their digital strategy playbooks and embrace LLM integrations with cautious optimism.
Cue the creatives: For those interested in the content creation use case, there is an integration to plugin Canva (an online graphic design platform) to your ChatGPT account — enabling end users to generate design templates and infographics using their OpenAI accounts. Due to high demand, the plugin is currently on a waitlist and only available to users subscribed to the Plus model.
What the LLM is doing here is streamlining the conceptual user experience for content creators by allowing them to quickly input prompt-based language to generate instantaneous visual inspiration via OpenAI’s latest GPT-4 model.
For those who grew up on dial-up internet and early versions of Microsoft Office (recalling memories of the animated Office Assistant character, Clippit), this is simply astounding progress.
Augmented Intelligence
As deep learning propels forward, and with a plethora of apps and plugins emerging for just about any productivity use case, how are organisations looking to integrate these tools for nuanced workflows? Moreover, how can human-centred design be applied to balance people and artificial intelligence in the future of work?
Current products based on LLMs such as OpenAI’s ChatGPT, Google’s Bard and Anthropic’s Claude may be cutting edge, however they can be inconsistent, providing so-called ‘hallucinations’ that can result in misleading information.
For enterprises with distinct data systems and complex knowledge architectures, this is a major concern. To address such drawbacks, companies can benefit from a modification technique known as retrieval-augmented generation (RAG).
Since LLMs are more attuned to reasoning engines that can generate inaccurate or static responses, RAG is a highly effective and cost-efficient way to store content for retrieval at generation time. In other words, applying RAG allows the deep learning model to fetch up-to-date or context-specific data from an external content store and then make it available to a LLM — reducing the risk of leaking sensitive data or providing misleading information.
IBM’s watsonx enables this functionality, claiming to ‘lower the computational and financial costs of running LLM-powered chatbots in an enterprise setting’.
Considering the future of work and the inevitable disruption to current jobs, augmented intelligence will be a driving factor to empowering performance and productivity through a more human-centric approach to technology.
To address existential concerns and by applying a more assistive intelligence automation approach, augmented intelligence offers a safer and more effective way to enhance human intelligence rather than replace it.
10 AI companies to watch
Keeping track of deep learning innovations is no easy task. Here are the top players pioneering this space.
Open AI: The non-profit research and development company with several LLMs (e.g. ChatGPT, GPT-4) and advanced image generation models (e.g. DALL-E) is in a state of flux. Since the restructuring of OpenAI in 2019, Microsoft has heavily backed the research organisation with a perpetual license to all of OpenAI’s intellectual property. However, the latest controversy surrounding Sam Altman and Greg Brockman’s departure from, and subsequent return to, the non-profit is stirring unrest about the rapid development of GenAI and commercialisation efforts. Can the non-profit still achieve its AI stewardship with its mantra for safe AGI for all? Expect OpenAI to reinforce its prosocial ambition in the years to come with closer ties to its founding mission.
Microsoft: The tech giant’s generative platform, Azure AI Studio, includes its Copilot portfolio (originally Bing Chat), which shows signs of significant potential and is primarily based on OpenAI technology. With a strong foothold in one of the world’s leading research laboratories, the multinational is well placed to dominate both B2B and B2C markets.
Google: With a dedicated research laboratory (DeepMind) and its content generator (Bard), plus additional generative products enabling organisations to build AI applications and explore LLMs via Google Cloud, the tech behemoth has a formidable integration scope — not to mention its guardian, Alphabet Inc. Expect its new conversational assistant, DuetAI, to rival ChatGPT and Copilot.
Apple: While the iOS provider has not been as vocal in this field, it plans to bring GenAI to all of its devices. Most notably, in-device AI, smarter apps and NLP are top on their agenda. Don’t underestimate this gadget Goliath’s machine learning capabilities.
Amazon AWS: In addition to the e-commerce provider’s Alexa app, new machine learning applications are offered with built-in GenAI via AWS. Expect Amazon to challenge its tech rivals with enhanced productivity and an intimidating AWS machine learning stack.
Baidu: The Chinese tech giant offers a full AI stack consisting of AI chips, deep learning framework, core machine learning capabilities, such as NLP, knowledge graph, speech recognition, computer vision and augmented reality. As one of the largest global AI companies, its cloud services business is bound for countless growth.
IBM: The ‘Big Blue’ multinational, primed to take on a sizeable portion of the cloud vertical, is investing big in its data analytics processing capabilities with solutions centred around its AI assistant, IBM watsonx.
Nvidia: The software and fabless company is a dominant provider of graphics processing units and hardware for powering various types of AI-enabled devices. Expect its supercomputing, platform software and machine learning models to revolutionise AI and data science.
Meta AI: Alongside its GenAI products (including its open source Llama 2 language model and the chatbot for generating text responses and photo-realistic images), the social meta verse company is investing big in its research laboratory aimed at improving augmented/artificial reality technologies.
Anthropic: This safety and research public-benefit corporation, founded by former members of OpenAI, is generating a lot of interest with its next-generation AI assistant, Claude — capable of conversational and text processing tasks while maintaining a high degree of reliability and predictability. Expect plenty of press about LLM steering and AGI in its foundational years.
Short courses
Wondering how to make sense of artificial neural networks? Need a more practical understanding of artificial intelligence tools and applications that will impact the way you or your team work?
Check out these two free courses by Andrew Ng from DeepLearning.AI. They provide a fun and insightful overview of AI with many real-world scenarios to wrap your mind around.
For a brief theoretical overview, Google Cloud’s learning path provides further context about the fundamentals and implications of GenAI. These courses are particularly useful in framing the big picture and can help inform next steps on how to position your career.
AI online courses from top institutions:
- Artificial Intelligence: Business Strategies and Applications (Berkley)
- Artificial Intelligence: Implications for Business Strategy (MIT)
- A Crash Course in AI (Stanford)
- Building an AI-Enabled Organization (Stanford)
- Transforming the User Experience through Artificial Intelligence (Stanford)
- Oxford Artificial Intelligence Programme (Oxford)
