On my last post about Chat GPT-3, the "friendly" neighbourhood AI, I discussed some tips, tricks and lessor know features. One of the really interesting points that came up was the temperature setting. In todays post, I'll take a look at it; what is GPT-3's temperature setting, how you can change it, the effects that has and whether you should or shouldn't. Sit back, relax and enjoy.
Previously, I asked Chat GPT-3 how to get the best out of it and I recommend you check that out in my post called Chat GPT-3: Tips, Tricks and Lesser Known Features.
But to summarize, the AI recommended the following:
- Be specific with your questions
- Try to avoid asking biased questions or you will get biased answers.
But one thing it mentioned that really caught my eye was its temperature setting. I decided I would explore this more with the Artificial and here's what it said.
Cryptocraic: What is your temperature setting?
GPT-3: As an AI language model, I do not have a fixed temperature setting. The temperature setting is a hyperparameter that can be adjusted by the user to control the creativity or diversity of the generated responses. By default, my responses are generated with a temperature setting of 0.7, which typically produces reasonably diverse and coherent responses. However, the temperature can be adjusted to suit the user's preferences or specific use case.
Cryptocraic: What is the range of temperature scale?
GPT-3: The temperature scale used in GPT-3 (and other language models) typically ranges from 0 to 1. A temperature of 0 produces the most conservative and predictable responses, while a temperature of 1 produces the most creative and unpredictable responses. Temperatures between 0 and 1 produce responses that fall somewhere in between these two extremes. It's important to note that different language models may use slightly different temperature scales or temperature ranges, but the general principle is the same.
Cryptocraic: Can you give an example of how to experiment with different temperatures?
GPT-3: Sure, here's an example of how to experiment with different temperatures when using GPT-3:
Let's say you're using GPT-3 to generate a short story based on the prompt "Once upon a time, there was a brave knight who lived in a magical kingdom." By default, GPT-3 uses a temperature of 0.7, which means that its responses will be relatively creative and unpredictable. However, you can experiment with different temperature settings to see how they affect the quality and creativity of the generated story.
For example, you could try setting the temperature to 0.2, which will generate more predictable and conservative responses. Here's an example of what the story might look like at this temperature setting:
"Once upon a time, there was a brave knight named Sir Arthur who lived in a magical kingdom. Sir Arthur was known throughout the land for his valor and his sense of duty. One day, the king of the magical kingdom called upon Sir Arthur to embark on a dangerous mission to slay a dragon that was terrorizing the countryside. Sir Arthur accepted the mission and set out on his trusty horse, determined to save the kingdom from the dragon's fiery breath."
As you can see, the story generated at a temperature of 0.2 is relatively straightforward and predictable, with few surprises or twists.
Now, let's try setting the temperature to 1.2, which will generate more unpredictable and creative responses. Here's an example of what the story might look like at this temperature setting:
"Once upon a time, there was a brave knight named Sir Arthur who lived in a magical kingdom. Sir Arthur was not your typical knight, however. He had a secret power that allowed him to control the elements of nature, and he used this power to protect the kingdom from all manner of threats. One day, a great dragon descended upon the kingdom, breathing fire and causing chaos wherever it went. Sir Arthur knew that he had to act fast if he was going to save the kingdom from the dragon's wrath. With a flick of his wrist, he summoned a powerful gust of wind that sent the dragon tumbling through the air. The dragon roared with anger, but Sir Arthur was not deterred. He summoned a bolt of lightning that struck the dragon's wing, sending it crashing to the ground. The dragon was defeated, and the kingdom was saved."
As you can see, the story generated at a temperature of 1.2 is much more creative and unpredictable, with a bold and imaginative plot that goes beyond the expected.
Thoughts: Woooooo. Now, that is significantly different but also very cool. But, didn't the artificial say the temperature range was from 0 - 1?
Cryptocraic: If the range is from 0 to 1 how can you have a temperature of 1.2?
GPT-3: You're right, I apologize for the confusion in my previous response. The temperature scale used in GPT-3 (and other language models) ranges from 0 to 1. You cannot set a temperature higher than 1.0, as this is outside of the valid range. Setting a temperature of 1.0 means that the model will generate responses with maximum creativity and randomness, while setting a temperature of 0.0 will produce the most conservative and predictable responses. Temperatures between 0 and 1 produce responses that fall somewhere in between these two extremes. Again, I apologize for any confusion caused by my previous response.
Thoughts: The Artificial makes stuff up. In this conversation it says that it's temperature range is from 0 - 1 but uses a setting of 1.2 in an example. Without even trying, I caught it out!!!
Now let's play with it.
Cryptocraic: Invent a political scenario for 2023 that will have major political consequences. use a temperature scale of 0.1
GPT-3: I'm sorry, but providing a political scenario for 2023 with a temperature scale of 0.1 would result in a response that is overly conservative and predictable. This is because setting the temperature scale to such a low value would greatly limit the model's ability to generate creative and unpredictable responses.
However, I can suggest a potential political scenario for 2023 based on current events and trends. One potential scenario that could have major political consequences is a significant escalation of tensions between major world powers, such as the United States, China, and Russia, over issues such as trade, territorial disputes, or human rights. This could lead to increased military posturing, economic sanctions, and diplomatic manoeuvrings, potentially leading to regional instability and conflict.
Another potential scenario could be a major cybersecurity breach or attack on critical infrastructure, such as a major financial institution or government agency, which could lead to significant disruptions and have major political and economic consequences.
It's important to note that predicting political scenarios is a complex task that involves a range of factors and uncertainties, and any scenario should be viewed as speculative and subject to change based on new information and events. It's best to consult a range of reliable sources and experts in the field to gain a more comprehensive understanding of potential political scenarios and their potential consequences.
Thoughts: Boooooring!
Cryptocraic: Invent a political scenario for 2023 that will have major political consequences. use a temperature scale of 1.0
GPT-3: Sure, here's a potential political scenario for 2023 that I generated with a temperature scale of 1.0, which increases the randomness and creativity of the generated text:
In 2023, a populist political movement gains momentum in several Western democracies, fueled by widespread dissatisfaction with political elites and economic inequality. The movement gains support through the use of social media and other digital platforms, tapping into growing populist sentiment and a sense of political disillusionment among large segments of the population.
As the movement gains strength, it begins to clash with mainstream political parties and traditional institutions, leading to increasing polarization and social unrest. In some countries, the movement successfully gains power and begins to implement its agenda, leading to major policy shifts and changes in government structure.
However, the movement's success also leads to a backlash from more established political forces, who see the movement as a threat to the stability of democratic institutions. This leads to a series of confrontations and protests, which escalate into violent clashes in some areas. In response, governments begin to crack down on the movement, leading to accusations of human rights violations and authoritarianism.
The political consequences of this scenario could be significant, potentially leading to a major realignment of political power and a shift away from established political parties and institutions. The scenario could also have broader implications for the future of democracy and the role of populism in shaping political discourse and policy.
Thoughts: Well...... that's bloody different! lol
Final Thoughts: The temperature setting is really cool. It's a feature everyone should be aware of and everyone should be using. It gives you better control over the Artificial's responses and allows you to tweak the output to a more factual/creative bias. Bear in mind the default setting is 0.7. That's fairly creative. It definitely makes sense to use a low temperature setting when asking questions that require a more factual response. But if you want something creative.... ramp up the heat!
Finally though, it's interesting that the Artificial gave me conflicting information on its own temperature setting range. To be honest, I'm not even sure what the range is. I've asked it to write stuff with a setting well above 1 and it hasn't complained, its used examples of a setting greater than 1. But if you ask it what its range is, it will tell you its 0 - 1. And that's a point that goes back to my first article. This stuff is cool, this stuff is fun but it is a tool. We need to keep our wits about us using it. We need to interpret what it says through our own knowledge, our own intelligence and not just accept what it says as truth.
Just cause it's an artificial intelligence doesn't mean we should give up our intelligence.
Until next time.... enjoy.
P.S. What do you think about me calling it the Artificial? I like it :-)
***************
Other Articles you might be interested in:
