InvestMacro

By Jonathan May, University of Southern California 

It doesn’t take much to get ChatGPT to make a factual mistake. My son is doing a report on U.S. presidents, so I figured I’d help him out by looking up a few biographies. I tried asking for a list of books about Abraham Lincoln and it did a pretty good job:

Number 4 isn’t right. Garry Wills famously wrote “Lincoln at Gettysburg,” and Lincoln himself wrote the Emancipation Proclamation, of course, but it’s not a bad start. Then I tried something harder, asking instead about the much more obscure William Henry Harrison, and it gamely provided a list, nearly all of which was wrong.

Numbers 4 and 5 are correct; the rest don’t exist or are not authored by those people. I repeated the exact same exercise and got slightly different results:

This time numbers 2 and 3 are correct and the other three are not actual books or not written by those authors. Number 4, “William Henry Harrison: His Life and Times” is a real book, but it’s by James A. Green, not by Robert Remini, a well-known historian of the Jacksonian age.

I called out the error and ChatGPT eagerly corrected itself and then confidently told me the book was in fact written by Gail Collins (who wrote a different Harrison biography), and then went on to say more about the book and about her. I finally revealed the truth and the machine was happy to run with my correction. Then I lied absurdly, saying during their first hundred days presidents have to write a biography of some former president, and ChatGPT called me out on it. I then lied subtly, incorrectly attributing authorship of the Harrison biography to historian and writer Paul C. Nagel, and it bought my lie.

When I asked ChatGPT if it was sure I was not lying, it claimed that it’s just an “AI language model” and doesn’t have the ability to verify accuracy. However it modified that claim by saying “I can only provide information based on the training data I have been provided, and it appears that the book ‘William Henry Harrison: His Life and Times’ was written by Paul C. Nagel and published in 1977.”

This is not true.

Words, not facts

It may seem from this interaction that ChatGPT was given a library of facts, including incorrect claims about authors and books. After all, ChatGPT’s maker, OpenAI, claims it trained the chatbot on “vast amounts of data from the internet written by humans.”

However, it was almost certainly not given the names of a bunch of made-up books about one of the most mediocre presidents. In a way, though, this false information is indeed based on its training data.

As a computer scientist, I often field complaints that reveal a common misconception about large language models like ChatGPT and its older brethren GPT3 and GPT2: that they are some kind of “super Googles,” or digital versions of a reference librarian, looking up answers to questions from some infinitely large library of facts, or smooshing together pastiches of stories and characters. They don’t do any of that – at least, they were not explicitly designed to.

Sounds good

A language model like ChatGPT, which is more formally known as a “generative pretrained transformer” (that’s what the G, P and T stand for), takes in the current conversation, forms a probability for all of the words in its vocabulary given that conversation, and then chooses one of them as the likely next word. Then it does that again, and again, and again, until it stops.

So it doesn’t have facts, per se. It just knows what word should come next. Put another way, ChatGPT doesn’t try to write sentences that are true. But it does try to write sentences that are plausible.

When talking privately to colleagues about ChatGPT, they often point out how many factually untrue statements it produces and dismiss it. To me, the idea that ChatGPT is a flawed data retrieval system is beside the point. People have been using Google for the past two and a half decades, after all. There’s a pretty good fact-finding service out there already.

In fact, the only way I was able to verify whether all those presidential book titles were accurate was by Googling and then verifying the results. My life would not be that much better if I got those facts in conversation, instead of the way I have been getting them for almost half of my life, by retrieving documents and then doing a critical analysis to see if I can trust the contents.

Improv partner

On the other hand, if I can talk to a bot that will give me plausible responses to things I say, it would be useful in situations where factual accuracy isn’t all that important. A few years ago a student and I tried to create an “improv bot,” one that would respond to whatever you said with a “yes, and” to keep the conversation going. We showed, in a paper, that our bot was better at “yes, and-ing” than other bots at the time, but in AI, two years is ancient history.

I tried out a dialogue with ChatGPT – a science fiction space explorer scenario – that is not unlike what you’d find in a typical improv class. ChatGPT is way better at “yes, and-ing” than what we did, but it didn’t really heighten the drama at all. I felt as if I was doing all the heavy lifting.

After a few tweaks I got it to be a little more involved, and at the end of the day I felt that it was a pretty good exercise for me, who hasn’t done much improv since I graduated from college over 20 years ago.

Sure, I wouldn’t want ChatGPT to appear on “Whose Line Is It Anyway?” and this is not a great “Star Trek” plot (though it’s still less problematic than “Code of Honor”), but how many times have you sat down to write something from scratch and found yourself terrified by the empty page in front of you? Starting with a bad first draft can break through writer’s block and get the creative juices flowing, and ChatGPT and large language models like it seem like the right tools to aid in these exercises.

And for a machine that is designed to produce strings of words that sound as good as possible in response to the words you give it – and not to provide you with information – that seems like the right use for the tool.

About the Author:

Jonathan May, Research Associate Professor of Computer Science, University of Southern California

This article is republished from The Conversation under a Creative Commons license. Read the original article.

By Jie Wang, UMass Lowell 

Empowered by artificial intelligence technologies, computers today can engage in convincing conversations with people, compose songs, paint paintings, play chess and go, and diagnose diseases, to name just a few examples of their technological prowess.

These successes could be taken to indicate that computation has no limits. To see if that’s the case, it’s important to understand what makes a computer powerful.

There are two aspects to a computer’s power: the number of operations its hardware can execute per second and the efficiency of the algorithms it runs. The hardware speed is limited by the laws of physics. Algorithms – basically sets of instructions – are written by humans and translated into a sequence of operations that computer hardware can execute. Even if a computer’s speed could reach the physical limit, computational hurdles remain due to the limits of algorithms.

These hurdles include problems that are impossible for computers to solve and problems that are theoretically solvable but in practice are beyond the capabilities of even the most powerful versions of today’s computers imaginable. Mathematicians and computer scientists attempt to determine whether a problem is solvable by trying them out on an imaginary machine.

An imaginary computing machine

The modern notion of an algorithm, known as a Turing machine, was formulated in 1936 by British mathematician Alan Turing. It’s an imaginary device that imitates how arithmetic calculations are carried out with a pencil on paper. The Turing machine is the template all computers today are based on.

To accommodate computations that would need more paper if done manually, the supply of imaginary paper in a Turing machine is assumed to be unlimited. This is equivalent to an imaginary limitless ribbon, or “tape,” of squares, each of which is either blank or contains one symbol.

The machine is controlled by a finite set of rules and starts on an initial sequence of symbols on the tape. The operations the machine can carry out are moving to a neighboring square, erasing a symbol and writing a symbol on a blank square. The machine computes by carrying out a sequence of these operations. When the machine finishes, or “halts,” the symbols remaining on the tape are the output or result.

Computing is often about decisions with yes or no answers. By analogy, a medical test (type of problem) checks if a patient’s specimen (an instance of the problem) has a certain disease indicator (yes or no answer). The instance, represented in a Turing machine in digital form, is the initial sequence of symbols.

A problem is considered “solvable” if a Turing machine can be designed that halts for every instance whether positive or negative and correctly determines which answer the instance yields.

Not every problem can be solved

Many problems are solvable using a Turing machine and therefore can be solved on a computer, while many others are not. For example, the domino problem, a variation of the tiling problem formulated by Chinese American mathematician Hao Wang in 1961, is not solvable.

The task is to use a set of dominoes to cover an entire grid and, following the rules of most dominoes games, matching the number of pips on the ends of abutting dominoes. It turns out that there is no algorithm that can start with a set of dominoes and determine whether or not the set will completely cover the grid.

Keeping it reasonable

A number of solvable problems can be solved by algorithms that halt in a reasonable amount of time. These “polynomial-time algorithms” are efficient algorithms, meaning it’s practical to use computers to solve instances of them.

Thousands of other solvable problems are not known to have polynomial-time algorithms, despite ongoing intensive efforts to find such algorithms. These include the Traveling Salesman Problem.

The Traveling Salesman Problem asks whether a set of points with some points directly connected, called a graph, has a path that starts from any point and goes through every other point exactly once, and comes back to the original point. Imagine that a salesman wants to find a route that passes all households in a neighborhood exactly once and returns to the starting point.

These problems, called NP-complete, were independently formulated and shown to exist in the early 1970s by two computer scientists, American Canadian Stephen Cook and Ukrainian American Leonid Levin. Cook, whose work came first, was awarded the 1982 Turing Award, the highest in computer science, for this work.

The cost of knowing exactly

The best-known algorithms for NP-complete problems are essentially searching for a solution from all possible answers. The Traveling Salesman Problem on a graph of a few hundred points would take years to run on a supercomputer. Such algorithms are inefficient, meaning there are no mathematical shortcuts.

Practical algorithms that address these problems in the real world can only offer approximations, though the approximations are improving. Whether there are efficient polynomial-time algorithms that can solve NP-complete problems is among the seven millennium open problems posted by the Clay Mathematics Institute at the turn of the 21st century, each carrying a prize of US$1 million.

Beyond Turing

Could there be a new form of computation beyond Turing’s framework? In 1982, American physicist Richard Feynman, a Nobel laureate, put forward the idea of computation based on quantum mechanics.

In 1995, Peter Shor, an American applied mathematician, presented a quantum algorithm to factor integers in polynomial time. Mathematicians believe that this is unsolvable by polynomial-time algorithms in Turing’s framework. Factoring an integer means finding a smaller integer greater than 1 that can divide the integer. For example, the integer 688,826,081 is divisible by a smaller integer 25,253, because 688,826,081 = 25,253 x 27,277.

A major algorithm called the RSA algorithm, widely used in securing network communications, is based on the computational difficulty of factoring large integers. Shor’s result suggests that quantum computing, should it become a reality, will change the landscape of cybersecurity.

Can a full-fledged quantum computer be built to factor integers and solve other problems? Some scientists believe it can be. Several groups of scientists around the world are working to build one, and some have already built small-scale quantum computers.

Nevertheless, like all novel technologies invented before, issues with quantum computation are almost certain to arise that would impose new limits.

About the Author:

Jie Wang, Professor of Computer Science, UMass Lowell

This article is republished from The Conversation under a Creative Commons license. Read the original article.