It helps to understand the field of AI holistically instead of getting distracted with the newest buzz like ChatGPT, DALL-E, and whatever new products will launch soon.
These new product launches (ChatGPT, DALL-E etc) may seem like AI innovation rapidly caught us by surprise. In fact, Cade Met author of Genius Maker, notes the start of modern AI space around early 1960s.
To holistically understand the AI space of today, it is essential to understand its 63 years of history with the lens of incentives – individuals who researched and advanced the field, the institutions who funded and hired talent for profit, advancement in technologies that advanced AI (computer vision etc), and governments and countries wanting to exploit AI for power and politics.
Let’s start with a brief history of the AI field.
Brief History of Artificial Intelligence
The development of AI field is roughly 63 years old – starting from the early 1960s. It is a story that starts from the research in academic institutions by a few bright minds. Soon after, AI becomes of interest by corporations to use it for profit and for politics, governments, and army to use it for power.
The first 50 years of AI field can be summarized by a few bright minds dedicated to advancing the research and frameworks for AI. These 50 years also saw almost 15 years of stalled research due to lack of funding or public interest. The most notable research individuals during these 50 years were – Jürgen Schmidhuber (ideas helped drive rise of deep learning), Dean Pomerleau (used neural network to build a driverless car during 1980s to 1990s), Joy Buolamwini (explored bias in face recognition services), Geoff Hinton (considered as founding father of deep learning movement) was at Google from 2013 to 2023.
In the past 23 years is when we see the rapid increase in competition and application of AI by companies like Facebook, Amazon, Google etc.
By 2012, it was widely recognized that conditions were ripe for the next phase of AI. Cost of data storage going down, GPUs, big corporation budgets, and domain of “deep learning” enabled an opportunity for profits and product improvements for those ready to exploit it. The few people who knew deep learning (the researchers mentioned above) were highly sought after. The race to acquire AI talent and setup their research labs was intense amongst the big companies. To attract talent, companies spent millions on state-of-the-art AI research labs. The founders Larry Page, Sergey Brin, and Mark Zuckerberg personally got involved and invited the talent for dinners at their homes. Every tactic was explored at any expense.
Having understood the history of AI, let’s look at current landscape where there is competition amongst companies and countries to advance and apply AI for their objectives (profits or power – or both).
Current competition in exploiting AI
Understanding the application and introduction of AI technology in everyday products is important in understanding the current landscape.
By 2012, the competition for voice assistants was hot. It served as a new medium and potential for commerce and consumption (eg. hey google! order me pizza). After, Apple’s Siri, google was hard at work to come out with a competing product and defend its market share. Later in 2012, google launched its on assistant, Google Now, and it surpassed Siri’s capabilities. In 2014, Amazon followed with Alexa. Soon after others came out with voice assistants too – Baidu’s DuerOS, Microsoft’s Cortana, and Facebook’s M.
The intense competition that followed among companies and countries can be understood through the lens of Google.
Google has been an early-mover and financially incentivized to exploit AI to advance and protect its profits and mission of “organize world’s information”. Most notably, in 2013, Google bought Geoff Hinton’s company, DNNResearch, to further advance its voice assistant/search and image search technology. Soon in 2014, Google spent $650 million to acquire DeepMind – research lab led by world leading AI researcher Demis Hassabis.
Later in 2014, Ilya Sutskever (now at OpenAI since 2016) working at Google AI lab developed the technology that enabled efficient language translation. In translation different sequences of words had to be related . That is, translation of sentences is much harder than individual words. For this, Sutskever work was “a breakthrough in any AI problem that involved a sequence”. After 18 months, the work by Sustkever was turned into Google Translate – a translation service that is used by millions and supports over 100 languages – the largest of its kind.
Google used its acquisition of DeepMind led by Demins Hassibis to develop AlphaGo – AI-driven who can play the game of Go. In 2016, AlphaGo beat the player of the decade Lee Sedol and beat the world champion Ke Jie in China in 2017.
After two months, in 2017, China announced its plan to be a world leader in AI by 2030.
Simultaneously, Google’s DeepMind Health was developing technology to read MRI, CT and eye scans diagnosis to improve the lives of doctors and patients in identifying the disease and getting additional insights from the vast historical data-set. They began working with NHS – the national heath service of UK. However, it was stopped by the British regulators for concerns regarding sharing of personal health information.
By 2018, Google released BERT, a “universal language model” or now known as LLM (Large Language Model) or simple LM (Language Model). BERT is being used to improve its search results – make it more accurate and relevant. These massive neural networks learn language from reading enormous amounts of text written by humans. A system using BERT was able to pass a Grade 12 science test.
In 2019, OpenAI, demonstrated a robotic hand that could solve Rubik’s Cube. The race to find robotic solutions to manual labour jobs such as warehousing and supply chain (influenced by amazon) was already picking up steam. Pieter Abbeel, OpenAI’s roboticist, left the company in 2018 and with his former students formed a new company, Covariant.
Most recently, OpenAI has launched two products that have caught the news and world by storm. ChatGPT was released by OpenAI which is a text-based conversational product that learned language by consuming a large data set. Before that, DALL-E, a deep learning model that generates images based on text-input. Shortly after, Bing came out with “the new bing” that you can chat with and partnered with OpenAI’s DALL-E. From China, Baidu launched Ernie Bot. Other competitors, like they did for voice assistants, will follow with their own product launches.
What is next in the field of AI
The field of AI will be advanced by its various players (Academics, Companies, and Governments) incentivized to push it further.
Academics and researchers will continue to explore the boundaries of AI – and attempt to get to AGI or any intellectual challenge that inspires them. We will see academics leaving or joining corporations to advance their research and intellectual challenges.
Companies will continue applying AI to advance their products to make them better for their their customers (such as google image search). The competition is intense and will continue to intensify to be the most innovative and best product for consumers and businesses.
Lastly, the world politics and powers, such as China, US and others will continue to exploit it for their advantage. Researching and applying AI to for military, surveillance, intelligence gathering and more.
Regardless, it will be defining chapter to watch as it will impact (even if its slowly over decades) all aspects of human life – power, politics, labour, profits, and more.