Stream enables you to rapidly gain a comprehensive understanding of any topic, even a topic as broad, complex and disruptive as artificial intelligence (AI). The full impact of artificial intelligence, in particular large language models (LLMs), is too much for one single blog post, indeed for a single human mind. We will focus on AI impact by industry to give you a sense of where adoption is high or low, and which applications are disruptive vs mostly hype. Even this humble corporate blog has already been disrupted, as the image was generated by DALL-E.
ChatGPT Already Replacing “Bread-and-Butter” Ad Agency Work
“The way I’ve used it right now, [ChatGPT] has the ability to generate copy. It has the ability to generate presentations. There’s one client I have, a tech guy there, he’s generating presentations on it. It doesn’t need to turn to an agency to help them put together some kind of proposals or something like that, marketing plans, these kinds of things. That kind of technology erodes or cuts into what agencies traditionally used to do. Agencies make a lot of money doing a lot of the little [unenjoyable] stuff, the bread-and-butter stuff. It’s not sexy but there’s a lot of money there. Once you take that away, it becomes more difficult.”
LLMs Can Interpret SEC Filings and Legal Documents
“If I’m a developer and I want to use OpenAI, I don’t have to train the model. The model is already pre-trained and it’s kept on the shelf and I can go and access it. The partnership runs, as I said, exclusive on Azure, all the training using billions of parameters.
What they have also done is they have democratized access to all these large language models because they’re very expensive and you need very important skill sets to optimize and train these models. Now, if I’m a small start-up or if I’m even a mid-sized company, I don’t need to have a very dedicated team. There are a lot of applications for it. It’s a very generic model. You could use it for a financial application like for example, one of the very promising applications I’m seeing is you can ask open OpenAI or ChatGPT to go through a lot of SEC filings. It will understand that. You can ask questions about any SEC filing or you could take any legal document as you may have read about and give it as an input. It will use this large language model to understand the legal document, organize its thoughts internally, and then you can ask questions and then it will give you answers about that document. It basically democratizes access to large language models to lots and lots of small and mid-sized companies who never could afford to do this. That’s why I think the announcement and this partnership is very instrumental.”
Industry Expert Thinks MSFT’s OpenAI Powered Bing Search Could Be a Disruptor in the Search Industry
AI Has Multiple Applications for Hearing Aids
“One of the companies that actually do AI currently is Widex. They have this SoundSense Learn where you can listen to different sounds and then it will optimize based on AI to your personal preferences. I think things like this and personal optimization of your hearing and your hearing aid settings is the thing that’s going to increase in the coming years.
Also, there’s a big issue within the hearing aid industry of addressing the markets where you don’t have so many trained professionals, so hearing care professionals and audiologists. Here AI automatic tools like chatbots can assist you in fitting your hearing aids and guiding you to self-fitting, and also being able to better calculate settings. That’s going to be an improvement and maybe also for sales, that’s perfect.
In the hearing aid itself, you can do machine learning because of the latency issues. Hearing aids are computers, but you have limited power and storage, you cannot do very, very advanced machinery within the hearing aid itself. It has to be connected to something, like a phone.”
LLMs Could Disrupt SNOW
“I agree with you that ultimately I could see a very disruptive world and maybe Databricks builds it or maybe Alteryx builds it or maybe a start-up builds it where someone just says, ‘All you need is our GPT-backed tool.’ You got to plug in your data to our tool, of course. We need the data to get the answers from. If you say, ‘This is my Databricks,’ or ‘This is my Snowflake, we’ll get you the answers,’ that to me seems like a future we’re moving towards.
Obviously everyone wants to be there. If you win that war, which is what Databricks did. Databricks was not a data storage company, they were a Spark company, but then Databricks said, ‘Hey, you’re running all of your jobs on Databricks. Why not use the Delta Lake storage technology?’ They realized that if they went deeper into the stack, they’d be harder to tear out and they’ve succeeded in that regard.
If you or I make a company tomorrow which just says, ‘Plug in Databricks, plug in Snowflake, and we will give you the best analytics experience ever,’ people will start using us, and then very quickly we could probably say, ‘Hey, we built our own cloud data warehouse and actually you could pipe your data directly into us and you don’t need to use Snowflake anymore.’ That to me seemed like the most obvious way that a Snowflake gets disrupted in the next 10 years.”
Oil & Gas Has High AI Adoption Rates Due to Regulatory Pressure
“There’s a lot of governmental pressures for more efficient production of fossil fuels. I think there’s a big component there. Obviously, safety is always top of mind for oil and gas. They’ve been high adopters of digital twins. AI is just the new component of that. Because they’re so sensitized and have so much data, they are faster for adoption as opposed to manufacturing where you could have a very large manufacturing facility but really very little on the centralization of the manufacturing line, depending on the manufacturing industry. That’s why oil and gas adoption was much higher.”
Former Competitor Believes AI Has Had Many Problems With Implementations and Scaling Solutions
AI Applicability in e-Commerce Depends on SKU Count
“I don’t think there’s a ton of value for [LLMs] for e-commerce. I think there is a little more about the findability aspect of things, because one problem that you run into, my experience in search is that your historical way of being able to make sure someone can find an item is you have to make sure that the description of that item matches what that customer is searching. If it doesn’t, you add what you call synonyms to make sure that synonyms to the words that are on the actual product to make sure that they match the words that customers are finding.
With the newest technology that has been coming out and been getting really popular in the last couple of years, like semantic search and some of these language models, is that you can actually be a lot more intelligent around how you’re matching products to customers interests than what you historically could. Now, how much is that going to move the needle? That’s more important for companies that have just millions and millions of SKUs, and less important for companies that are focusing on just maybe a few hundred SKUs.”
AI Tools Still Rudimentary in Hotels
“[AI is] very nascent. It’s very early stages. I’ve had many companies talk about how they’re using AI to help them figure out rates or how to make more money if there’s a sale going on or a big conference coming to town. It seems like it’s still very rudimentary. It still needs you to put in the information and it’ll spit something out.”
This was a lot of information, on a wide range of topics, but this is just a small sample of what’s available from Stream on artificial intelligence and large language models. For access to the full expert transcript library, request a free trial here.
Content Marketing for Stream by AlphaSense
Austin’s primary experience is in consulting and private equity, though he’s also a published author.
Read all posts written by Austin Moorhead
