• Ben's Bites
  • Posts
  • Canva becomes the AI magician, can you?

Canva becomes the AI magician, can you?

PLUS: AI evaluation, Assistant with Bard and Perplexity API

Sign up | Advertise | Ben’s Bites News
Daily Digest #256

Hello folks, here’s what we have today;

PICKS

  1. Canva launches Magic Studio, partners with Runway ML for video. Canva is integrating AI into all parts of its platform to automate tasks and inspire designs. 🍿Our Summary (also below)

  2. Assistant with Bard - At the Made by Google launch event, Google teased the upcoming combination of Google Assistant and Bard. 🍿Our Summary

  3. Anthropic's primer on challenges in AI evaluation. The AI safety team at Anthropic outlines challenges in different techniques of evaluation across multiple dimensions like accuracy, bias, and harm. 🍿Our Summary (also below)

  4. Perplexity launches pplx-api - Its new, low latency API for accessing open-source LLMs like Mistral 7B, Llama2 13B, Code Llama 34B, and Llama2 70B.🍿Our Summary (also below)

  5. The rise of the novice coder - Can AI turn every employee into a developer?

TOP TOOLS

  • Moonvalley - Create breathtaking cinematic & animated videos from simple text prompts.

  • Intently - Turn LinkedIn actions into sales opportunities.

  • Melon - Bring your second brain to life.

  • Poe API v2 - Use answers from any Poe bot as input in your workflows.

  • Coda 4.0 with Coda AI - The connected work assistant that knows your company.

  • AI hype by Replicate - All the latest Python repos in an HN-like list.

  • Snipd - Blinkist for podcasts. 5 min summaries of your favourite podcasts.

  • WebLangchain - An open-source web research assistant with citations. (blog)

  • Metaphor Researcher - Use Metaphor to find high-quality content and create a summary.

  • Turbo Puffer - A truly serverless vector database.

NEWS
QUICK BITES

Canva is the place where most of the workforce goes for their day to day design work. They’ve been silently adding AI image generation features that suit those existing needs under the “Magic” suite of tools. Now they’ve come with bang, adding more features, and expanding to video generation by partnering with Runway.

What is going on here?

Canva is launching Magic Studio, integrating AI into all parts of its platform to automate tasks and inspire designs.

What does this mean?

Here are the highlights of launched features under the new Magic Studio:

  • Magic Design - Turn ideas into designs instantly with AI-generated templates.

  • Magic Switch - Transform content into different formats and languages with one click.

  • Magic Grab - Make images editable like Canva templates for easy editing.

  • Magic Expand - Use AI to expand images beyond the original frame.

  • Magic Morph - Transform text and shapes with creative effects and prompts.

  • Magic Edit - Make complex image edits using simple text prompts.

  • Magic Media - Generate professional photos, videos and artworks from text prompts.

  • Magic Animate - Add animated transitions and motion to designs instantly.

  • Magic Write - Generate draft text and summaries powered by AI.

The partnership with Runway uses Runway’s Gen-2 text-to-video model to power the Magic Media features. Canva’s App Marketplace allows access to partner AI apps like DALL-E, Imagen and more.

With features, Canva is staying responsible with Canva Shield - their Built-in safety, privacy and security controls for AI features and Creator Compensation - a $200M fund to pay creators who opt-in for training AI models.

Why should I care?

Art usecases aside, most people don’t see image generation fit in their workflow. Canva is capturing this by modifying the models to their target audience’s need i.e. making design and images, often in a professional space, quickly without learning heavy tools.

This levels the playing field in design. Novices can make pro-level content quickly without technical skills. The AI features automate the busywork like finding a relevant stock image, recreating in different sizes and more so you can focus on communicating your idea.

Unclassifieds - short, sponsored links

  • Pragma.ai - An AI that can use the tools you use like Notion and Slack - Try it for free.

  • infoGPT - Unleash AI Editor & Art Studio, audio notes, chat with PDF, boost writing & productivity with all-in-one digital AI assistant.

QUICK BITES

Perplexity launches pplx-api, its new, fast and efficient API for accessing open-source large language models like Mistral 7B, Llama2 13B, Code Llama 34B, and Llama2 70B.

What is going on here?

pplx-api enables blazingly fast LLM inference, up to 2-3x lower latency than competing solutions.

What does this mean?

The key benefit is ease of use. Developers can harness powerful language models without needing deep expertise in deploying and managing them. Costs are another big factor. For Perplexity, themselves, migrating one core feature from external APIs to pplx-api cut costs by ~4x for the company.

The API is proven in production at scale, reliably serving ~1 billion tokens daily. With a roadmap covering custom models, embeddings, and more, pplx-api is positioned to be a go-to solution as demand for industrial-grade LLM inference grows. Data transmitted via the API is automatically deleted after 30 days and Perplexity does not train on user data.

Why should I care?

With pplx-api, you can easily integrate cutting-edge LLMs into their applications with a familiar REST API, without needing to manage complex model deployment and inference infrastructure. pplx-api abstracts away this complexity through an optimized system leveraging NVIDIA TensortRT-LLM on AWS infrastructure. Low latency and cost savings are factors to look at if you are already using any API from other open source LLM aggregators.

QUICK BITES

The AI safety team at Anthropic outlines the challenges in AI evaluation across multiple dimensions like accuracy, bias, and harm. They share what they have learned from extensive experience developing and implementing standardized tests and human evaluations of AI systems.

What is going on here?

Anthropic explain evaluating AI is complex; even basic benchmarks require extensive engineering and hide pitfalls.


What does this mean?

From multiple-choice tests to human ratings, Anthropic has found every evaluation approach has major limitations.

Small changes can swing scores for multiple choice evals which makes them easy to cheat. More complex evaluations like tests for bias and harm take months to properly implement, and hide their own flaws until you dig deeper.

Real-world human evaluations of harm and social impacts are expensive, subjective and legally risky. Third parties build great frameworks, but adapting our models for them takes non-trivial effort. More collaborating between companies and evaluators would help, but balancing objectivity with expertise is tricky.

Why should I care?

As AI becomes more capable, having trustworthy ways to evaluate progress and safety will be crucial for developing beneficial systems. However, current evaluation methods are limited, so we can't fully rely on them.

AI developers, policymakers, and companies using AI must recognize evaluations' shortcomings today. More funding and coordination are needed to improve them. With thoughtful research and engineering, we can build the trustworthy, unbiased assessments essential for deploying AI safely at scale.

Ben’s Bites Insights

We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below;

  • All 10k+ links we’ve covered, easily filterable (1 referral)

  • 6k+ AI company funding rounds from Jan 2022, including investors, amounts, stage etc (3 referrals)

Join the conversation

or to participate.