“So we were ushered into a room and the president came in. He was engaged, he was witty…he was asking really good questions. Each of us had a chance to talk to him about various ideas & one idea that was raised was the issue of extreme misinformation, particularly around the election cycle.”

That’s one of the stories from Oren Etzioni, technical director at the AI2 Incubator, on this episode of Shift AI, a show that explores what it takes to thrive and adapt to the changing workplace in the digital age of remote work and AI. 

We discuss Oren’s background and experience incubating AI companies, and we get his take on his recent private conversation with President Biden about policies that will shape the future of AI.

Listen below, and continue reading for highlights from his comments, edited for context and clarity. Subscribe to ShiftAI and hear more episodes at ShiftAIPodcast.com. 

First paying job: My first job and still one of my most favorite was when I was in high school. I had a part-time job as an ice cream scooper at Baskin-Robbins, and I was paid minimum wage. The job did have one major benefit, I was allowed to eat as much ice cream as I could possibly consume. So I ate a lot of ice cream and made very creative milkshakes. It was a phenomenal job. 

Family and upbringing: Both of my parents were professors, and they very much encouraged me to go into the family business, which was being an academic. My parents are very intellectual, so we often had conversations about issues of the day; regarding economics and political elites. The topics my parents, and I discussed most were in the realms of political science, sociology, and policy and as a result, it was very natural for me to choose a topic and delve into it deeply. 

AI in the Workplace: I look at the macro outlook for AI in the workplace, as making meetings more productive, and automating tasks with provision. While we’re ruminating about the future of work, it’s way more than that, we’re building something that’s going to make work a lot better for individuals. 

Artificial Intelligence can serve as a dynamic tool to amplify productivity within the workplace. I believe two Seattle-area startups, Read AI and Yoodli, epitomize this vision completely.

• Read.ai uses AI to monitor engagement & mood within video meetings, in addition to automatic meeting summaries
  
• Yoodli wields AI as a creative tool for writing, focusing on text generation and language translation. 

These tools are complementary in assisting the user in areas of reading and writing creativity. I have a dream that one day, these tools will be combined into a singular entity of an AI assistant: a tool that looks over your shoulder and guides you in reading and writing. This is potentially a powerful tool that could also exist inside of glasses in an augmented reality setting. 

Point Solutions vs Incumbents: First, It is true that the incumbents have something of an advantage, as it requires a lot of work to build and train these models.

At the same time, it’s rapidly changing with models like LLAMA2, and other open source LLMs. There’s a very interesting intersection between the models created by the major incumbents and cloud providers as you have these open source models that are very widely available and I think that balance of power is still to be determined. 

Second thing that’s happened is, there’s been a tremendous amount of companies “putting lipstick on an API” by taking one of these very powerful APIs and building a small interface around it. These simple point solutions are often not enough, so most of these companies focus on a multitude of API-related solutions. 

There’s a lot of opportunities to build more sophisticated tools for the workplace and the home that utilize generative AI but do it in more powerful ways. 

Limitations around Large Language Models: A mundane limitation is around latency and cost. If you’re trying to use these models through the API in real time, it can take a while. Something like search can be irritating and the costs can add up based on the size of the task. 

A more profound issue around artificial intelligence is that you can’t 100% rely on the information that comes back from these models. They can be toxic, and they can refuse to answer based off of platform guardrails.

The biggest problem is that an LLM can hallucinate and it can double-down on the wrong information, presenting it as facts. The model is only thinking about completing the sentence, not about presenting what is factual and you have to be very careful verifying information.  

AI Ethical Considerations: There are numerous ethical issues when it comes to artificial intelligence like privacy, the impact on democracy, and surveillance. You have countries like China using technology that we thought would enhance democracy, yet it’s actually furthering a totalitarian hold on power. 

I think there’s also a major issue around jobs, and it could happen a lot faster than people think. It’s the kind of thing we want to see the government intervene appropriately through training programs, and making sure people are treated appropriately. 

I’m still very optimistic that generative AI can be an empowering tool; an enabler for increased productivity which is ultimately good for everybody, driving a lot of economic growth. We just need to ensure the weakest members of society are treated right, as productivity is increased, and wealth is created. 

During his private meeting with President Biden: An idea that came out of chatting with President Biden was the issue of extreme misinformation, particularly around the election cycle. A flood of misinformation is very hard to identify and stop before it makes an impact on the election. Our elections have been so close, it could really tip the election in one direction, or another. I’m very interested in figuring out how to leverage some of the research and tools to build a coherent response to some of these events. 

Another topic we talked about was how American society invests its resources for moonshot AI projects. Something generally that we as Americans love is moonshot projects and the president is very passionate about making a bold move to address things like cancer and thinks AI has a tremendous role to play in revolutionizing biology. 

Watch or listen to the full conversation above, and find more episodes here.

Read More