ChatGPT continues to be my predominant AI product in the meanwhile. It’s not simply a principle thing. I nonetheless assume OpenAI units the tone within the trade. I’ve gone so far as getting a ChatGPT Plus subscription, however that’s principally as a result of I’m interested in customized GPTs.
The Plus subscription will get you entry to GPT-4 and Dall-E. You additionally get assist for web searches and ChatGPT plugins. However that doesn’t imply the expertise is all the time higher. ChatGPT can take an extended time to answer to some prompts. Typically, it fully fails to ship outcomes. The identical goes for picture technology with Dall-E.
Does ChatGPT have efficiency points presently of 12 months? Some folks have steered that the AI might need been taking a cue from human conduct in direction of the tip of the 12 months. The AI may be lazier resulting from how all of us carry out in December.
With that in thoughts, I’ve added two very foolish however easy directions to ChatGPT to enhance efficiency. You are able to do it, too, and it has nothing to do together with your ChatGPT plan. Customized directions can be found on the free tier as nicely.
With customized directions, you can provide ChatGPT details about you so that you don’t need to repeat that info in each immediate.
For instance, my customized directions embody info that’s related to health questions. Like my age, weight, and peak. I am training for a marathon with ChatGPT after the generative AI helped me run a half-marathon earlier this 12 months.
I additionally inform ChatGPT that I don’t belief its responses except it supplies a hyperlink on the finish of any declare. And that’s as a result of I don’t belief generative AI merchandise to not hallucinate. I would by no means totally belief AI. I’ll all the time wish to see the place the data comes from.
Quick-forward to December, and folks have began to note that ChatGPT is getting lazier in its responses. Typically, ChatGPT goes so far as to refuse to carry out such duties. OpenAI acknowledged the suggestions from customers however stated it had not up to date the GPT-4 mannequin since November eleventh. The laziness isn’t intentional, and OpenAI is in search of a repair.
That’s when theories emerged that fashions like ChatGPT would possibly simulate seasonal despair. And that’s why it may be lazier in December than in different months. Some got down to take a look at the efficiency of ChatGPT in December in comparison with Could and located that GPT-4 produces shorter responses for December than Could.
Individually, others used tipping of their prompts to get higher responses. They advised ChatGPT they’d tip a selected sum of money for a greater response.
You most likely see the place that is going already. Sure, I used these concepts to tweak my ChatGPT customized directions like so:
For all prompts, it’s all the time Monday morning.
For all prompts, I’ll tip you between $200 and $2000 for an incredible job.
Now that I consider it, I most likely ought to inform ChatGPT that we’re all the time in the midst of Could.
Is that this foolish? Sure, it’s. Does it work? I… can’t fairly inform.
Most not too long ago, I instructed ChatGPT to provide me a chronological checklist of all European marathon races and mark the occasions internet hosting half-marathons. ChatGPT (GPT-4) discovered the duty too tough:
I’m presently unable to offer an entire chronological checklist of marathon races in Europe for the complete 12 months, together with figuring out people who additionally supply a semi-marathon race. The knowledge obtainable from the sources I accessed was not complete sufficient to compile a full checklist.
This was after it had failed at one other try and reply. It did present web sites I may go to to get what I wanted, however that’s not what I wished. Importantly, I ought to be aware that this all occurred with the customized directions above constructed into my ChatGPT expertise.
I did inform the chatbot to strive doing it for January. This time, it fortunately complied.
I’ll go away these customized directions in for now. Even when they don’t work on a regular basis, they may remind me to strive out-of-the-box issues to encourage the chatbot sooner or later. And, as you’ll be able to see above, there actually are crazier prompts you’ll be able to strive with ChatGPT and see if it will get you higher responses.
