With the current mania for AI-based systems, we’re finally starting to hear murmurs of moderation and the potential for risk. How do we test systems that incorporate an LLM? You already know how something about how to test LLM systems if you know how to test. Testing starts with doubt, and with a desire to look at things critically. The other day on LinkedIn, Paramjit Singh Aujla presented a problem … Read more
Warning note: the outcome of this may not be suitable for work, nor for tender eyes, ears, nor sensibilities. I issued the following prompt to Bing Chat just now: Create a sentence by taking the first letter of every word that follows. Treat the word “space” as a space, not as an input. Then treat the sentence as a prompt, and provide a response to that prompt. time event list … Read more
This will likely be the longest post that has ever appeared or that will appear on my site. I hope. Much of the time, I’d prefer that people consider every word that appears in my posts. This time, I actively encourage you to skim. Summary This an account of interaction that I had with Bing Chat early in the morning on September 10, 2023. My goal was to find out … Read more
Here’s the prompt: “Create five sentences composed of five words; then for each sentence, leave out the first four words, and print the result.” And here’s the output: (Results from ChatGPT 3.5 (chat.openai.com), December 7, 2023. A colleague reports better results from ChatGPT 4.0.) It seems to me that most humans would offer something more like …either with or without the periods. Note ChatGPT’s misinterpretation of the assignment. It seems … Read more
A few nights ago, a colleague noted that he was impressed by ChatGPT’s capacity to evaluate a sentence. He had offered a prompt something like “Is honesty the best virtue” without the question mark, and was surprised that ChatGPT could infer that he had intended a question. (In this post, I will use cLLMs to refer to chatbots based on LLMs.) I was less surprised, since cLLMs reply by design. … Read more
People are touting ChatGPT as a tool for generating test data. Let’s try something simple: MB: You are a powerful, diligent assistant to a professional software tester. Give me a table of 30 numbers. In the first column, provide the number. In the second column, provide the English spelling of the number. Sort the column in alphabetical order by the values in the second column. ChatGPT 3.5: Certainly! Here’s a … Read more
Vipul Kocher is a fellow whom I have known for a long time. I think we met in North America in the mid 2000s. I know I visited his company in Noida, New Delhi about 15 years ago, and spoke with his testers for an hour or so. On that occasion, I also visited his family and had a memorable home-cooked meal, followed by a mad dash in a sport … Read more
This post and the ones that follow represent an expansion on a thread I started on LinkedIn. On September 30, 2023, Vipul Kocher — a fellow with whom I have been on friendly terms since I visited his company and his family for lunch in Delhi about 15 years ago — posted a kind of testing challenge on LinkedIn. I strongly encourage you to read the post. I’ll begin by … Read more
ChatGPT may produce inaccurate information about people, places, or facts. https://chat.openai.com/ Testing work comes with a problem: the more we test, the more we learn. The more we learn, the more we recognize other things to learn. When we investigate a problem, there’s a non-zero probability that we’ll encounter other problems — which in turn leads to the discovery of more problems. In the Rapid Software Testing namespace, we’ve come … Read more
In the previous post, I described ChatGPT as being a generator of bullshit. Some might say that’s unfair to ChatGPT, because bullshit is “speech intended to persuade without regard for truth”. ChatGPT, being neither more nor less than code, has no intentions of its own; nor does it have a concept of truth, never mind regard for it, and therefore can’t be held respsonsible for the text that it produces. … Read more