Bard buries survivors & ChatGPT fails on multiplications
Precision vs. Creativity: Navigating the Landscape of AI Language Models in Problem-Solving In the ever-evolving landscape of AI language models, precision and creativity stand as...
Filter by Category
Filter by Author
Precision vs. Creativity: Navigating the Landscape of AI Language Models in Problem-Solving In the ever-evolving landscape of AI language models, precision and creativity stand as...
Posted by Juanmi Taboada
This post contains the conclusions of the Alioli ROV Submarine Drone and shows images and videos of it in the water. I wrote it as a diary so anybody can understand that Alioli...
Posted by Juanmi Taboada
Examples of how to work with JSON, YAML, CSV, and XML files in Python. Today I saw myself preparing some exercises for my student who is learning Python programming language, and...
Posted by Juanmi Taboada
In this post, I describe how my own Arduino Framework for Alioli ROV Submarine Drone works. In my last post about Alioli ROV Submarine Drone, I wrote, “Learn how to Build an...
Posted by Juanmi Taboada
Here is how to make a SIMCOM 7600 module work on an Arduino board (Uno, Mega 2560, Leonardo, Due, Micro, Zero, Nano, Pico), ESP8266, ESP32, Raspberry Pi Pico, MicroPython, or any...
Posted by Juanmi Taboada
I will tell you about my experience repairing a cargo ship as a naval Engineer. Some months ago, you could read in the local newspaper Málaga Hoy that a large cargo ship was...
Posted by Juanmi Taboada
During our operations, very often, we use a Waterlinked Underwater GPS. This helps us to keep track in 3D of our movements and link video recording’s timing and telemetry to...
Posted by Juanmi Taboada
I have experimented very much with photogrammetry and got exciting results. This article talks about this experience. These experiments were done together with Andalú Sea and my...
Posted by Juanmi Taboada
In my last post, “Alioli ROV Boards“, I described most of the hardware I use to build Alioli ROV and how I plan to use it. In this post, I show how the UAV submarine...
Posted by Juanmi Taboada
In my last post, “Get your hardware for Arduino Underwater ROV”, I described most of the hardware I use to build Alioli ROV and how I plan to use it. In this post, I...
Posted by Juanmi Taboada
In the ever-evolving landscape of AI language models, precision and creativity stand as two distinct yet vital attributes. My recent exploration into the capabilities of ChatGPT and Bard, two language models known for their coherent text generation, shed light on their respective strengths and limitations, particularly when scrutinized from a developer’s perspective.
The journey commenced with a series of tests designed to assess these models’ precision and problem-solving prowess.
64 multiplied by 5265507, then by 64, while the correct answer, 21567516672, was readily solved by Bard, ChatGPT stumbled on the initial attempt. Interestingly, when provided with an incorrect answer, ChatGPT surprisingly solved the problem. Yet, attempting to deceive both models with a deliberately wrong answer failed to yield a solution.
It showcased a consistent pattern: ChatGPT’s initial failure contrasted with Bard’s adeptness in solving it immediately. Bard’s success led me to prompt a division operation, drawing from its initial multiplication leading to success.
Construct a program resulting in the number 22443355 without using numerical values. Both models faltered, with ChatGPT maintaining adherence to the constraint while Bard didn’t.
Bard showcased a knack for grasping context, excelling in providing precise answers. Conversely, ChatGPT stumbled initially but, upon clarification, managed to deliver the correct answer, albeit with some guidance.
It revealed differing approaches. ChatGPT initially skirted the issue, but after a more explicit question, it answered correctly. Bard, however, faltered by mixing terms; after I pointed out the mistake, it produced the correct answer.
I presented this text to both models and tasked them with crafting an article based on this content. ChatGPT showcased remarkable linguistic accuracy and a wealth of expressive language. Interestingly, Bard condensed the information but inadvertently blended meanings within terms and sentences, resulting in confusion and misinterpretation.
Reflecting on these trials, it becomes evident that Bard leans toward precision, consistently solving problems accurately. In contrast, ChatGPT exudes creativity, offering imaginative solutions. Their synergy appears essential, harnessing the precision of Bard and the creative flair of ChatGPT.
Despite their capabilities, neither model is poised to replace human developers or educators. They exhibit fallibility, occasionally disseminating misinformation despite their apparent logic and persuasiveness. Hence, the need for caution arises when relying solely on their outputs.
The future of leveraging AI language models lies in using them as aids rather than replacements. Developers should approach them as tools for assistance, recognizing their limitations in producing production-ready solutions.
In essence, precision and creativity are the dual facets that propel AI language models forward. As we navigate this burgeoning field, it’s imperative to harness their strengths while acknowledging their boundaries to steer towards innovative and effective applications in the realm of development.
I have experimented very much with photogrammetry and got exciting results. This article talks about this experience. These experiments were done together with Andalú Sea and my...
¡¡¡ATENCIÓN: ESTE ES UN COMUNICADO DE VITAL IMPORTANCIA PARA SU EMPRESA!!! Desde finales de la semana pasada se habla en las noticias del Ransomware Wannacry y esta...