The world of AI has been shaken by Google’s dismissal of AI Ethicist Dr Timnit Gebru last week. This behaviour is emblematic of the self-centred attitudes of major tech companies which also results in lack of commitment to democratisation of technology. With Facebook, Amazon and Apple also in the spotlight for allegedly creating a monopoly, the time has come for SMEs to re-revaluate their AI providers.
UNSUSTAINABLE GROWTH
The tech industry has experienced a meteoric rise this millennium, growing into one of the world’s largest industries, with investment increasing by £3.1 billion in 2019 alone. Amazon, Apple, Facebook and Google have a combined worth of $4 trillion, giving them unprecedented power over the marketplaces they facilitate. This affords them enormous control as they set the rules that other businesses operate under. In this way, these companies take responsibility for democratising the access to AI and enabling advancements in this field. However, can these companies be trusted to do it?
The recent dismissal of Dr Timnit Gebru, has put Google’s motivations into question. Gebru is a computer scientist renowned for her work into the racial and gender biases in facial recognition systems. She is also known for being the co-founder of Black in AI, an organisation championing black female voices in the AI community. Gebru cited the reason for her termination as censorship of an attempt to publish an article outlining the potential ethical issues that may arise from Google’s use of AI. MIT review of the research paper in question reveals that the controversy was raised around CO2 emissions of processing the large AI models such as BERT, Google’s search engine. This calls into question Googles commitment to develop ethical Artificial Intelligence, instead prioritising unsustainable business practices.
‘CHOKE COMPETITION’
This attitude towards promoting self-interest over ethics has been apparent over the course of the US House of Representatives Antitrust hearings that have also engulfed Facebook, Amazon and Apple. Each has been accused of practises bordering on a monopoly, with Facebook being accused of ‘buying and freezing out small startups to choke competition.’
Facebook’s acquisitions of WhatsApp and Instagram have been accused of anticompetitive conduct. The committee presented evidence of Mark Zuckerberg stating that “moving faster and copying apps, can prevent our competitors from getting footholds”. The FTC’s Bureau of Competition Director, Ian Conner, has stated “Facebook’s actions to entrench and maintain its monopoly deny consumers the benefits of competition”.
Apple is currently embroiled in a lawsuit from Epic Games where its restrictions on third-party purchases for IOS apps are deemed as anti-competitive. They maintain that the lack of multiple marketplaces on Apple devices inhibits businesses from making competitive profits.
Amazon has come under fire, as Jeff Bezos stated in July’s Antitrust Hearing that he cannot promise that Amazon hasn’t broken policy to access independent seller data to inform their competitive business. This is particularly showcased by its poor treatment of these independent sellers allegedly referring to them as “internal competitors”. The hearing also brought to light a dependency by software development companies on Amazon services such as AWS. These businesses are therefore almost forced to use a cloud service which cannot guarantee the protection of their data.
HOW TO RETAIN CONTROL OF YOUR BUSINESS TECHNOLOGY
These cases do not outright present the businesses as untrustworthy, but certainly suggest their primary objectives are self-interested. Allegations of a monopoly aside, the level of control over the market by the major tech companies is readily apparent. Without democratisation of tech and AI, SME’s become reliant on services from major tech companies, these services are then essential to the core processes and infrastructure of organisations, leading to these organisations having a lack of control within their business.
A lack of full control over processes is an inherent issue of subscribing for off-the-shelf AI solutions. Industries are adapting to the changes brought by Covid-19 by adopting rapid digital transformation or risk central processes becoming outdated and ineffective. Automating core systems with Artificial Intelligence, whilst a necessity, creates dependencies on third parties for businesses lacking an inhouse data science team.
There are of course serious arguments for subscription based software being an appealing choice for businesses seeking to embrace digitalisation. The speed of deployment, cost of investment and lack of maintenance costs for off the shelf solutions are examples of the attractive qualities of third parties. Indeed, bespoke AI is only as effective as stakeholder’s ability to identify and express their business problems and desired solutions.
NO SOFTWARE IS INFALLIBLE
The major difference between 3rd party and built solutions is the level of control businesses have over the software. Despite the initial drawbacks of developing one’s own systems, the long-term benefits far outweigh the short-term gains from outsourcing.
Anyone with experience of technology will understand that at some point it is likely to fail on you. From spending hours on the phone to customer services to fix your online banking to the struggle of connecting your laptop to your printer, no technology is infallible. Just Monday of this week Google’s Docs, email services and YouTube were knocked offline for more than half an hour.
Apply this near guarantee to Artificial Intelligence and suddenly the issue is vastly inflated. It is not a matter of a piece of software break or fail to work, rather one in which your business core business decisions maybe inaccurately informed. Recognising the dangers relating to failure of AI systems the recent report by the UK’s Centre for Data Ethics and Innovation calls for a “a mandatory transparency requirement on all public sector organisations using algorithms that have a significant influence on significant decisions affecting individuals”.
Without this transparency one cannot understand why an AI system is failing. Given the apparent attitudes of the major tech companies there is no evidence to suggest they will offer this kind of visibility over the decision making in their products. Without a customer centric approach from the big four, businesses are in danger of falling by the wayside as the only AI they have access to fails to provide for them. There is no customer support line for data bias or problematic output. The only way for business to gain technology which works for them is to retain control over systems.
BUILD YOUR OWN AI AND STAY IN CONTROL
With competition being swallowed up by the big tech companies, it seems the only means of developing long term trustworthy systems is to build them bespoke.
The core difficulty with developing bespoke Artificial Intelligence is the lack of democratisation of knowledge and access to the technology. This manifests itself in staffing requirements, as top-level talent moves to big tech or academia, or in technological understanding, wherein a business cannot bring together stake holders to develop effective systems.
Finding good talent is the backbone to any AI project. The complexity of the technology necessitates highly specialised talents across different disciplines of AI, machine learning, data science, data engineering and DevOps. Further difficulties are added to this puzzle as stakeholder find it difficult to relay project ambitions to PhD-level data scientists. Effectively road mapping projects is essential to understanding staffing needs and project requirements. In the current market businesses will clearly benefit from bringing in AI expert consultants to plan their projects. Support in strategy development of bespoke AI solutions gives business the clarity and control they need to regain confidence in the undeniable power of Artificial Intelligence technologies.
originally posted on forbes.com by Kasia Borowska
Author’s Statement: I concentrate on cutting-edge Artificial Intelligence & Machine Learning applications. I am the cofounder and CEO of Brainpool.ai, a worldwide network of 300 AI and Machine Learning experts. I hold an MSc in Cognitive and Decision Sciences from UCL and a BSc in Mathematics from Greenwich University, UK. I write about topics such as True AI, democratizing AI, AI ethics, applications of AI and ML in business and how to bridge the gap between AI academia and business. I focus particularly on financial, marketing, healthcare and retail verticals but will report on any businesses where AI and ML impacts.
