Designer Trained Her Own AI to Make Art

Understanding AI as a Creative Tool

Introduction to the Discussion

• The speaker discusses receiving significant backlash online, including death threats, for their views on AI in design.

• They introduce Esha Vera, a designer at Perplexity who utilizes AI tools like Midjourney and emphasizes that AI is part of a broader creative process rather than the final output.

• Esha has faced similar online hostility but continues to produce impactful work and share her experiences openly.

The Role of AI in Creativity

• The conversation focuses on viewing AI as a collaborator in the creative process instead of a replacement for human creativity.

• The podcast aims to explore what drives innovative designers and builders in today's digital landscape.

Personal Journey and Creative Expression

• The host expresses admiration for Esha's diverse creative endeavors, including her record label, and seeks insights into her journey.

• Esha reflects on her early experiences with platforms like MySpace and Neopets that sparked her interest in coding and design.

Early Influences on Creativity

• Both speakers reminisce about their formative years on different internet platforms that allowed self-expression through customization and coding.

• Esha shares how expressing herself online was foundational to her identity as a creator, leading to challenges when transitioning into professional work.

Balancing Professional Work with Personal Expression

• Esha discusses the importance of solving problems creatively while maintaining personal expression through various projects beyond traditional apps or websites.

• She highlights how she inadvertently took on roles such as creative director by collaborating with others to bring ideas to life.

Evolution of Design Roles

The Changing Landscape of Design

• The speaker reflects on the evolution of their role, noting that traditional definitions of a designer feel insufficient due to the diverse interests and outputs they now engage with.

• They express a reluctance to elaborate on their title as "designer," fearing it may pigeonhole them into a specific niche.

Collaborative Environment at Perplexity

• The speaker highlights the talent within the Perplexity team, emphasizing collaboration and idea pitching as key components of their work culture.

• They describe how working outside predefined boxes is encouraged, leading to innovative ideas being embraced by the team.

Culture of Trust and Autonomy

• A strong culture characterized by "high trust, high ownership, high autonomy" is noted as unique compared to previous tech experiences.

• Designers are empowered to address issues directly without waiting for formal processes; this includes fixing bugs collaboratively.

High Agency in Team Dynamics

• The speaker discusses building a culture of high agency within their team, where individuals own specific areas while collaborating freely.

• They seek tangible examples from Perplexity that illustrate how high agency aligns with company mission goals.

Collaboration and Knowledge Sharing

• Each designer has specialized knowledge but must collaborate across teams to leverage historical insights effectively.

• Regular communication with colleagues is emphasized as essential for successful project outcomes.

Innovative Use of Technology in Design

Generating Unique Invitations

• The discussion shifts towards a collaborative project involving creating special invites for public release, initially starting with QR codes that lacked uniqueness.

Embracing AI in Design Processes

• The speaker advocates for using AI creatively in design workflows, viewing it as an enhancement tool rather than just an output generator.

Image Generation Techniques and Tools

Creating Unique Assets

• The speaker discusses the process of combining brand imagery with Midjourney styles to create unique assets, utilizing Photoshop for composition.

• They trained several models using curated data, resulting in three distinct models that allowed for random element swapping during runtime.

• Each generated card is unique, featuring various elements like black holes and life streams, emphasizing intentional design choices to ensure character in each generation.

Scaling Image Generation

• The speaker reflects on the challenge of scaling image generation while maintaining uniqueness, considering tools like Floropana and Weeavy for visualization.

• They mention experimenting with different methods but found some approaches impractical due to complexity and the need for hand curation.

• A flexible approach was adopted to allow for imperfections in generated images, which were acceptable within the context of motion and style.

Tool Utilization and Training Data

• Midjourney is highlighted as a preferred tool due to its ability to produce unique shading through style prompting.

• The speaker references using AI Sevit (though uncertain about the name), along with Fall AI for generating images efficiently while controlling parameters like generation speed.

• They achieved rapid generation times while ensuring continuity across outputs by fine-tuning model runs.

Exploring New Tools

• The discussion shifts towards exploring new tools; designers often face challenges assessing feasibility at scale before deep diving into a method or tool.

• The speaker emphasizes a process involving iterative testing—evaluating outputs not just as final products but as part of an ongoing creative cycle.

Iterative Design Process

• Their methodology includes extensive back-and-forth between inputting images into tools and refining them through Photoshop collages or texture creation.

• This iterative approach allows them to integrate new tools into their existing workflows rather than relying solely on one tool's capabilities.

• They view experimentation as essential, spending time understanding how new tools can complement their established processes.

Exploring AI Prompting Techniques

The Importance of a Prompt Library

• The speaker emphasizes the value of maintaining a library of prompts in Midjourney, which has been built up since early beta usage. This library aids in understanding how to effectively write prompts.

Experimentation with Prompts

• Various styles have been tested for prompts, including poetry, abstract concepts, lyrics, and short stories. The speaker enjoys pushing the boundaries of what tools can generate.

Navigating Restrictions

• Due to stricter guidelines from Midjourney over time, older prompts that may contain undesirable content still work if used verbatim. However, any modifications often lead to flags or rejections.

Creative Workarounds

• The speaker discusses finding alternative ways to describe sensitive content (e.g., using "red liquid" instead of "blood") to achieve desired visual effects without triggering restrictions.

Communication as a Design Skill

• Effective communication is highlighted as crucial for designers today. It involves articulating ideas clearly to colleagues and engineers while also navigating AI tools successfully.

Challenges in Media Generation

Structuring Prompts for Better Outputs

• There’s an increasing trend among media generators using structured formats like JSON for more accurate outputs. The speaker reflects on their own experiences with prompt structuring.

Trial and Error in AI Interaction

• A personal anecdote illustrates the trial-and-error process when prompting AI tools; initial vague requests often fail until more detailed instructions are provided.

Complexity in User Engagement

• When testing an AI tool named Comet, the speaker had to provide extensive details about actions needed on a webpage due to hidden elements causing failures in execution.

The Future of AI Tools

Current Limitations of AI Technology

• The speaker notes that current AI technology is still developing and requires users to engage deeply with prompting techniques—often leading to frustration among non-experts.

Integration Over Prompting Focus

• Reflecting on their experience at Dcript, the speaker argues that integrating AI features into tools without requiring users to understand underlying prompts will enhance usability and adoption rates.

Moving Beyond Traditional Prompting

• Emphasizing user-friendly design, the speaker believes reducing reliance on explicit prompting will help demystify AI for users who may be hesitant or intimidated by it.

Exploring AI Integration in Creative Work

The Impact of Tinkering with AI

• The implementation and integration of AI into products significantly affect their perceived impact. Encouraging more people to experiment with AI could shift opinions positively.

Understanding Different Approaches to Prompts

• Advocating for a hands-on approach, the speaker emphasizes that trying out different paths with AI can lead to better understanding and acceptance.

The Science of Prompting

• Tatiana Siguva, a creative ambassador for Perplexity, shares her extensive catalog of prompts and parameters, highlighting the scientific aspect of crafting effective prompts.

Varied Approaches to Project Development

• The speaker discusses contrasting methods for project initiation—spending significant time on context versus starting with simple, spontaneous ideas.

Contextual Relevance in Prompting

• There is uncertainty about which prompting method yields better results; context plays a crucial role depending on the type of project (coding vs. media).

Streamlined Communication with AI

• The speaker often begins with informal, stream-of-consciousness prompts that resonate well due to the contextual understanding built over time.

Value in Contextual Inference

• Using Comet allows for deeper contextual inference beyond standard search engines, enhancing the quality and relevance of generated content.

Browsers as Foundations for Future Systems

• The discussion suggests that browsers may serve as foundational elements leading towards more advanced operating systems in AI development.

Balancing Creativity and Technology

Reconciling Pro-AI Stance with Self-Expression

• Despite being involved in tech design for AI companies, the speaker argues against the notion that this work stifles creativity or self-expression.

Personal Workflow Insights

• The speaker's workflow involves significant input into training data creation; they believe unique outputs cannot be achieved without this foundational effort.

Elevating Others Through Collaboration

• A key motivation is elevating others' talents by using AI-generated concepts as starting points rather than final products.

Collaborative Creation Process

• After generating initial ideas using tools like Sono, the speaker collaborates with producers to refine these concepts into unique works rather than releasing them directly.

This structured summary captures essential insights from the transcript while providing timestamps for easy reference.

Creative Collaboration and AI Tools

The Role of Creative Direction

• The speaker discusses their role as a creative director within a media studio, emphasizing collaboration with talented individuals to enhance creativity.

Communication Enhancement through Generative AI

• Generative AI tools have significantly improved the speed of idea communication, countering criticisms that they produce low-quality outputs.

• The concept of "slop" generated by AI is viewed as beneficial, providing a tangible basis for meaningful discussions and shared vocabulary among collaborators.

Human Curation in AI Outputs

• The importance of human curation in the generative process is highlighted; even if the final output isn't curated, prior curation ensures quality and relevance.

• A personal anecdote about meeting Kenneth, a creative director for notable games, illustrates how seeing respected peers use AI positively influenced the speaker's perspective on its ethical use.

Ethical Considerations in AI Usage

• Initially skeptical about AI due to ethical concerns regarding voice cloning technology at Dscript, the speaker emphasizes the need for responsible usage protocols.

• Comfort with using generative tools arose from ensuring that ethical safeguards were built into their applications to prevent misuse.

Evolution of Perception Towards Generative Art

• After experimenting with Midjourney, the speaker experienced an "unlock," realizing that generative art could complement their skills and creativity despite initial reservations.

• Comparisons are drawn between contemporary generative art and past controversies surrounding Photoshop's capabilities in altering images.

Navigating Public Perception and Criticism

• The speaker faced significant backlash on social media when sharing work involving generative art but sought to engage constructively with critics to understand their concerns.

• Over time, it became clear that ethical usage depends on individual intent; distinguishing between genuine artistic expression versus opportunistic content creation is crucial.

Current Landscape of AI-generated Content

• There’s an observation that much current content lacks depth or taste; many creators seem focused solely on gaining online attention rather than producing meaningful work.

• Acknowledgment that discerning what is authentic versus generated has become increasingly challenging in today's digital landscape.

Understanding AI: Navigating Questions and Ethics

The Importance of Inquiry in AI

• Emphasizes the need to ask questions about AI rather than harboring negative feelings towards it. Key inquiries include its creation process, purpose, and underlying motivations.

• Highlights a framework for understanding AI by focusing on its applications and the intentions behind its use, promoting curiosity over judgment.

Polarization Around AI Usage

• Discusses the divisive nature of opinions surrounding AI, noting that initial reactions can be extreme as people engage with new technologies like Midjourney.

• Reflects on personal experiences with ethical discussions in AI, particularly observing how Kenneth's innovative use of tools influenced perceptions.

Creative Applications of AI

• Observes Kenneth's boundary-pushing approach in using AI creatively while advocating for more flexible policies regarding content generation.

• Notes that the extent to which something is created by AI (10% vs. 100%) matters significantly in discussions about authenticity and creativity.

Transparency in Production Processes

• Raises questions about how much to disclose regarding the use of AI in media production, balancing human effort with technological assistance.

• Shares a commitment to valuing handcrafted elements within productions despite utilizing modern tools, emphasizing a blend of traditional and contemporary methods.

Personal Experiences with Disclosure

• Describes a personal journey toward transparency about using AI, initially providing disclaimers but later feeling that this led to increased negativity online.

• Conveys a shift away from constant disclosure due to backlash while still maintaining an open dialogue about the role of AI in creative work.

Art and AI: Personal Expression vs. Public Sharing

The Motivation Behind Art Creation

• The speaker emphasizes that their art is primarily for personal satisfaction rather than for online sharing, although they do share to educate others about possibilities in art.

Navigating Polarization in Art

• There’s a discussion on the current polarized environment, suggesting that artists may need to take breaks from consuming divisive content while still appreciating the work of others.

Tools and Techniques in Artistic Hobbies

• The conversation shifts to tools used in creative processes, with the speaker expressing interest in how different tools help conceptualize ideas, particularly mentioning their enjoyment of ink collection.

Daily AI Tools Utilized at Perplexity

• The speaker mentions using Midjourney and Craya as primary AI tools but acknowledges wanting to explore Cursor more extensively. They find value in observing how colleagues utilize these tools effectively.

Collaborative Creativity with AI

• Meetings often involve real-time collaboration where team members demonstrate their use of AI tools, enhancing creativity through immediate application during discussions.

Prototyping with Perplexity Labs

Quick Prototyping Solutions

• An example is given where the speaker created a simple onboarding page using Perplexity Labs, demonstrating how quick solutions can lead to effective prototypes that are later implemented into production.

Evolving Brand Guidelines

• Discussion includes modern brand books needing to incorporate prompt parameters alongside traditional elements like color schemes and patterns for better guidance on generative design.

Exploring Hobbies Through AI

Curiosity Driving Tool Usage

• The speaker shares their curiosity about new AI tools like Sunno, noting rapid advancements that inspire them to integrate these innovations into their creative practices.

Inspiration from Past Ideas

• They reflect on revisiting an idea from high school using current technology, showcasing how evolving tools can bring old concepts back into relevance and inspire collaboration with professionals.

Tattoo Culture and Personal Style

Trusting Tattoo Artists

• The speaker discusses their approach to tattoos by trusting artists whose styles they admire without needing a fully developed concept beforehand. This highlights a collaborative trust between artist and client.

Defining Artistic Styles

• A specific tattoo style is identified as black work with negative space elements. The speaker appreciates this aesthetic and values the unique artistry behind it.

Reflections on Design Evolution

Future of Design Skills

• Concluding thoughts emphasize that being an effective designer will increasingly depend on adaptability and integration of new technologies like AI into traditional design practices.

Communicating with AI: A New Collaboration

The Role of Prompting in AI Interaction

• Effective communication extends beyond clients and engineers to include the tools themselves, reframing prompting as a vital skill rather than a gimmick. This involves clearly articulating ideas, understanding constraints, and iterating based on feedback while collaborating with AI.

Understanding Credit in AI Creation

• The speaker emphasizes an important perspective on AI usage: acknowledging that while the output is not solely theirs, credit should not be entirely given away either. This includes recognizing the numerous micro-decisions made prior to engaging with the machine.

• Key elements contributing to successful AI collaboration include curating training data, building compositions using manual tools, and applying personal taste throughout the creative process.

Blurring Lines Between Using and Creating with AI

• There exists a blurred line between merely using AI and actively creating with it. Those who can navigate this distinction without over-explaining or apologizing are likely to produce work that holds significant value—referred to as "anti-slop."

• ESH exemplifies this approach by effectively integrating their creative processes with AI capabilities, leading to impactful outcomes.