{"id":50091,"date":"2023-04-18T08:45:13","date_gmt":"2023-04-18T13:45:13","guid":{"rendered":"https:\/\/simfoni.com\/?p=50091"},"modified":"2023-04-18T10:34:33","modified_gmt":"2023-04-18T15:34:33","slug":"ai-in-the-age-of-chatgpt","status":"publish","type":"post","link":"https:\/\/simfoni.com\/engineering\/ai-in-the-age-of-chatgpt\/","title":{"rendered":"Beyond the Hype: Managing Our Expectations of AI in the Age of ChatGPT"},"content":{"rendered":"
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t

\n\t\t\t\t“We overestimate the impact of technology in the short-term and underestimate the effect in the long run.”\t\t\t<\/p>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t\t\t\t\t\t\t- Ray Amara, Stanford computer scientist <\/cite>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\tTweet<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/blockquote>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Simfoni’s<\/span><\/a> Chief Technology Officer Alan Buxton<\/a><\/span><\/strong> shares insights into expectations around <\/span>the capabilit<\/span>ies<\/span> of AI<\/span> and what it means for <\/span>procurement<\/a>.<\/span><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

ChatGPT<\/span><\/a> has <\/span>brought ‘Artificial Intelligence’<\/strong><\/span> and specifically, ‘Generative AI’<\/strong><\/span> well <\/span><\/span>i<\/span><\/span><\/span>nto the mainstream<\/span>.<\/span> <\/span> This is a world-changing time<\/span>.  <\/span>(<\/span>It must have felt like this when the printing press came along<\/span>!)<\/span> It’s<\/span> impressive how quickly Large Language Models and Generative AI have caught the world’s imagination. <\/span><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t

\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

At <\/span>Simfoni<\/span>, <\/span>we’ve<\/span> been working intensively with <\/span><\/span>transformers<\/span><\/span><\/a>—<\/span>the technology that underpins <\/span>ChatGPT<\/span>,<\/span> since the middle of 2021 and have become familiar with both the strengths and pitfalls of this technology<\/span>.  <\/span>As we<\/span> mov<\/span>e<\/span> at pace through <\/span>Gartner’s Hype Cycle,<\/span> I want to share some insights <\/span>on the current “hype”<\/strong> surrounding <\/span>AI <\/span>to<\/span> help you navigate up <\/span>the <\/span>peak of inflated expectations (where we are now)—and through the <\/span>inevitable trough of disillusionment that is coming<\/span>. As exciting and transformative as AI may seem, navigating it in the world of procurement<\/strong><\/a> requires grounded understanding to avoid falling into disappointment<\/span>.<\/span><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

“AI in sourcing<\/a> and procurement <\/span>remains<\/span> overhyped, with use cases throughout the S2P<\/a> process. Most use cases are still <\/span>emerging<\/span>, but reliability on and adoption of longer-standing use cases is increasing. Maturity is highly variable depending on the use case, which is reflected across AP invoice automation (APIA), conversational UI, autonomous <\/span>sourcing<\/a><\/span> and autonomous procurement<\/span><\/a>.”<\/span><\/em><\/p>

Source: Gartner Hype Cycle <\/span>for<\/span> Procurement & Sourcing Solutions, 2022. By Analyst(s): Kaitlynn Sommers, Micky Keck, Patrick Connaughton, Lynne Phelan<\/span><\/span><\/strong><\/span> <\/span><\/p><\/blockquote>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

To begin, <\/span>l<\/span>et’s<\/span> be clear about <\/span>what <\/span>people mean when they say “AI<\/strong><\/span>”<\/strong>.<\/span> What most people think of AI is <\/span>very different<\/span> from what people researching or working in AI know it to be. Close your eyes for a few moments and see what comes to mind when you think of “AI<\/span>”.<\/span>  Did you think of something out of a sci-fi film with a conscious machine that looks and acts like a human but <\/span>isn’t<\/span>? <\/span>Bad news<\/span> if you did: <\/span>That’s<\/span> not what the industry has in mind when they say “AI<\/strong><\/span>”<\/strong>.<\/span> That’s<\/span> what we call “Artificial General Intelligence<\/strong><\/span>”<\/strong>,<\/span> or AGI. If I had my way, we would replace ‘AI’ with a more <\/span>accurate<\/span> term like ‘Assistive Computation’ <\/span><\/span>as proposed by <\/span>Ilyaz<\/span> Nasrullah<\/span><\/span><\/a>. This would <\/span>eliminate<\/span> a lot of confusion around the scope, <\/span>function,<\/span> and value of AI in procurement<\/a>.<\/strong><\/span><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

Read more: Guide to AI in Procurement<\/span><\/a><\/strong><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t\t

What is ‘AI’ to AI Professionals?<\/h2>\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Are you familiar with Pac-Man? The video game with the round yellow creature that <\/span>must<\/span> eat the dots while avoiding brightly-<\/span>colored<\/span> ghosts? Those ghosts are AI. Yep, you read that right. The ghosts follow simple rules about when to chase Pac-Man, which is <\/span>essentially AI<\/span>. (You can read all about the clever AI <\/span><\/span>algorithms that power them here<\/span><\/span><\/a>)<\/span>.<\/span><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

To researchers in the field, AI is something that <\/span>operates<\/span> independently to achieve an <\/span>objective<\/span>. This could be anything from catching Pac-Man to predicting stock quantity fluctuations. <\/span>That’s<\/span> it. There is no implication of knowing or understanding anything other than the task at hand. <\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

Keep this in the back of your mind as we come to <\/span>OpenAI’s<\/span> tool, <\/span>ChatGPT<\/span>. You can think of <\/span>ChatGPT<\/span> as the combination of two bits of tech: a chat engine, and the GPT language model. GPT is a proprietary language model created by <\/span>OpenAI<\/span>. Other companies like Meta, Google, Nvidia, and Microsoft each have similar language models. <\/span>All<\/span> these language models are fed vast quantities of text to <\/span>identify<\/span> patterns and predict what word will come next. Think of them as <\/span>auto-complete<\/span> on steroids<\/span>.<\/span> <\/span> <\/span><\/span> <\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

The<\/span> “T” in GPT is for “<\/span>T<\/span>ransformers<\/span><\/strong>”,<\/span> whic<\/span>h <\/span>expands<\/span> on traditional machine learning <\/span>methodologies to allow for a broader range of understanding and context<\/span>.  <\/span>Tra<\/span>n<\/span>sformers are what make GP<\/span>T models so advanced and effective at generating natur<\/span>al-sounding language<\/span>, and <\/span>ultimately what<\/span> led to <\/span>OpenAI’s<\/span> breakthrough.<\/span> Transformers<\/span> enhanc<\/span>e the<\/span> GPT language model by training it to create text that reads and feels more human-like<\/span> through a <\/span>technique called “Reinforcement Learning through Human Feedback” or RLHF<\/strong>. <\/span>(<\/span>Basically, having human input helps the algorithm learn what text looks natural and which <\/span>doesn’t<\/span>)<\/span>. Then, <\/span>OpenAI<\/span> added a Chat interface on top, making it easier to interact with the language model. And the result is <\/span>ChatGPT<\/span> – an impressively intelligent AI tool that can both generate text on a wide variety of topics and hold convincing conversations with humans.<\/span><\/span> <\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

So, here we have tech (in the tradition of the Pac-Man ghosts) that has identified patterns in written text and can generate human-like text with which you can converse.<\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\"PAC-MAN\"\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Now, let’s de-mystify its capabilities.<\/span><\/span><\/span><\/strong><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t\t\t\t\t\t

Does ChatGPT really “know” anything? Arguably, it knows what human-like text looks like. But it doesn’t know anything about what it is writing. It is optimized to generate content that is structured in a way that looks human. It is not optimized for creating anything true.  <\/span> <\/span><\/p>

What does this mean for you? I’m going to borrow from a <\/span>recent paper on the use of ChatGPT in Environmental Research<\/span><\/a>, which has a good summary of the pros and cons of ChatGPT.  They found that ChatGPT’s main advantages are:<\/span> <\/span><\/p>