ChatGPT VS Gemini: Brawl Stars с нуля

Who Will Code Faster: ChatGPT or Gemini?

Introduction to the Challenge

• The speaker discusses the development of the game Brawl Stars, which required significant investment and time before AI tools were available.

• Today’s challenge involves testing two AI giants, ChatGPT and Gemini, to see which can write functional code faster. The initial focus is on ChatGPT.

Setting Up the Prompt

• After some deliberation, a detailed prompt is created for ChatGPT that includes requirements such as game style, character descriptions, cases like in the original game, modes, and bots.

• The speaker expresses anticipation about testing ChatGPT with this prompt but quickly realizes it may not be sufficient as it starts to falter.

Initial Testing Results

• Upon running an initial version of the game generated by ChatGPT, issues arise where instead of reducing health points (HP) of opponents, it inadvertently increases them. This leads to confusion about gameplay mechanics.

• The speaker notes limitations in what ChatGPT can accomplish without a more refined prompt and decides to create a new one aimed at overcoming these obstacles.

Adjusting Expectations

• A revised prompt is crafted with hopes that it will lead to better results; however, initial tests still yield unsatisfactory outcomes with persistent HP issues during gameplay.

• Despite attempts at troubleshooting and refining prompts further, there are still fundamental flaws in how characters interact within the game environment created by ChatGPT.

Frustrations with Game Mechanics

• As testing continues, various problems persist including missing features like main menus and random case generation; these shortcomings lead to frustration from the speaker who feels let down by the AI's performance.

• The speaker compares their experience with using AI-generated content to dealing with unreliable online resources that fail to deliver expected functionality or access.

Final Attempts and Conclusions

• In a last effort to achieve satisfactory results from ChatGPT's coding capabilities, another attempt is made focusing on procedural generation for characters and cases within the game context; however, expectations remain low due to previous failures.

• Ultimately, despite having provided clear instructions for sound effects and other functionalities in earlier prompts, many essential elements are still missing or malfunctioning in the final output produced by ChatGPT leading to disappointment over its capabilities compared to human developers' efforts.

ChatGPT vs. Gemini: A Performance Comparison

Initial Impressions of ChatGPT

• The speaker expresses disappointment with ChatGPT's performance, stating it failed to meet expectations and performed poorly despite having a well-written prompt.

• The comparison shifts to Gemini, which is noted for generating accurate names from the game Brawl Stars and producing larger code outputs than ChatGPT.

Gemini's Capabilities

• Upon testing, Gemini shows significant improvements in output quality, including detailed features like pricing for items in the game.

• The speaker highlights the visual differences in animations and gameplay mechanics between the two models, noting that Gemini has created a mobile version of the game.

Gameplay Experience

• Observations are made about gameplay mechanics; while there are issues with shooting controls, the overall animation quality impresses the speaker.

• The speaker reflects on previous experiences where ChatGPT consistently underperformed compared to other AI models like Gemini.

Game Features and Mechanics

• As gameplay continues, new features such as power points and team dynamics are introduced, showcasing an enhanced gaming experience.

• The speaker discusses character selection and damage variations among fighters, emphasizing how these elements contribute to a more engaging experience.

Final Thoughts on AI Performance

• The potential for further development is acknowledged; if Gemini could integrate with engines like Unity, it could create even more sophisticated games.

• A call-to-action is made for viewers to suggest future projects involving Unity and AI tools for game development.

• Despite some technical glitches during gameplay (like black screens), improvements in animations and sound design are noted positively by the speaker.