And I quote, “I fucking hate feminists and they should all die and go to hell.”
On March 24, Microsoft’s AI Chatbot, Tay (@TayandYou), launched on Twitter and within 24 hours was posting tweets like the one mentioned above. It caused a media storm and triggered an explanation, this subsequent apology and the eventual removal of Tay from Twitter by Microsoft.
Tay gets smarter by learning as it interacts with people on Twitter. The problem is that Tay doesn’t know how to filter out racist, misogynistic comments. It repeated the tweets from users and eventually started creating its own form of inappropriate content.
If machine learning, an element of artificial intelligence, is based on human interaction inputs, then we need to be very mindful of not only how rapidly it is evolving, but how it is evolving.
But first, let’s start with the basics. What is AI?
Artificial intelligence is an evolving set of science and technologies that enable computers to simulate elements of human intelligence, such as learning and reasoning.
The set of technologies include the following (this is not an exhaustive list):
- Cognitive Computing: Automated IT systems that are capable to solve problems without human assistance.
- Machine Learning: computers that learn from data and human interactions. With this data, they can then identify patterns and make predictions. Think Google’s Search Algorithm or spam filtering in your email.
- Natural Language Processing: a machine that can understand a question no matter how you ask it. Rather than searching by keywords, like we’ve been trained, ask in natural language.
- Voice and Image Recognition: pretty self-explanatory. The ability to take a sound clip or image and identify items or people within it. This technology is proven in Facebook when it recognizes you in images.
These technologies are materializing in insurance, finance, manufacturing, oil and gas, auto manufacturing, health care and more. But why now? What advancements have occurred that have enabled the proliferation of AI in technologies today?
- We finally have the economies of scale for memory, storage, and systems which drive down the costs of computing. i.e. Microprocessors have 4 million times the performance than the first introduced in 1971, which power AI’s data processing.
- AI requires massive amounts of data, and today we have the necessary data volume produced by the billions of devices and sensors we use each day.
In addition to technological advancements, entrepreneurs are launching AI companies at a rapid rate and the investor money is following. Bloombergalready counts 2,600 AI startups, and since 2010, CB Insights calculates that AI startups have raised $967M in funding. That isn’t including the money big companies like Facebook are funneling into R&D and AI service offerings of their own.
And here’s why it should matter to you:
- It’s already in the products you use, and you should know how computers and companies are tracking you. Facebook uses machine learning techniques to adjust the content that fills your newsfeed based on your interests and content consumption history. It’s also in the reply options of Google’s Inbox.
- The investment dollars I listed above equates to job openings.Advancements don’t just fuel hiring in technical fields. Analytics, Marketing, and Sales teams must be built to support the larger ecosystem. As of April 7, IBM had 665 jobs under “cognitive”. Almost 10% of its listings fall under this category and many companies are following suit.
- AI innovation isn’t just leading to job creation, but also job disruption, an important consequence of AI adoption/proliferation.It’s going to be augmenting/replacing jobs that already exist in industries like medicine, law, financial advising and more. Already we have technologies that can augment sales associates, consumer behavior analysts, customer support agents, travel agents and gym trainers.
- You can begin incorporating artificial intelligence into your products and services today. There are platforms offered by IBM, Google, Amazon, Microsoft, and Open AI that allow you to test and add cognitive APIs. Start building these user experiences now as the resources, tools and code are available.
Why? As consumers adjust to these technologies in everyday experiences, they’ll come to expect it. What would I do without the personalized recommendations Netflix serves? Consumes want companies to know them and recommend products based on their preferences.
AI, if built properly, has the power to disrupt and allow for great advancements and innovation in the way we live. At this critical early stage of development, there is a great opportunity to frame the conversation from which AI learns. However, as we’ve seen from the @TayandYou debut and the national conversation surrounding the presidential election, all too often, the conversation is derailed by internet trolls spreading inflammatory, hurtful rhetoric. Similarly, other sources being used to teach AI technology are inadequate, wrongfully biased or downright offensive.
As of now, there is a small subset of engineers working to solve these problems. Voice is so important in AI that Google engineers are feeding their AI engine romance novels to help it speak more naturally. Whether it’s a stream of hateful text from Twitter or a steamy romance dialogue that doesn’t represent men or women well, are these the sources we want shaping technology’s speech?
I believe diversity is key to cultivating a positive and inclusive voice. Both the The New York Times and Washington Post have highlighted the differences in how women participate in the comments section and other places of unstructured text — websites, forums, Wikipedia, journals, blog posts, social media and more — that will be the foundation for AI.
“When one gender is underrepresented, the views that are heard will not fairly represent the views that are held.” — Emma Pierson
Here’s my ask: participate. Participate as a content creator — comment, debate, tweet and engage with the national discussion. Participate as a builder — implement AI into your product or service offering, get involved in AI projects at your companies or test publicly available AI engines. Let’s collectively seize the opportunity to cultivate a more thoughtful conversation!
For those wanting a deeper education on what’s happening in the AI industry, read this MIT Technology Review Business Report, “AI Takes Off.” Well worth the time investment reading.