August 2023 has been a busy month for advancements in artificial intelligence, marking the end of the Summer season. In this blog post, we aim to give life science marketers a concise overview of the important developments and insights that have taken place.
The insights presented in this blog post are gathered from the conversations featured in episode 20, episode 21, episode 22 and episode 23 of the "Artificially Intelligent Marketing" podcast, hosted by our CEO, Paul Avery, along with his industry peer, Martin Broadhurst.
For those who would rather listen, episodes 20, 21, 22 and 23 are available for streaming below (or you can subscribe to the podcast on your platform of choice).
Summary of episode 20:
Protecting Your Content from AI Bots
OpenAI has provided a way to block its chatbot from accessing your website content without permission. This helps safeguard proprietary docs and user-generated content from being scraped for training generative AI models built by OpenAI. Using the provided crawler details, webmasters can update their robots.txt files and firewall rules to restrict the bot. For digital marketers, it enables more control over what content is indexed and used for AI training.
OpenAI's ChatGPT User Growth and GPT-4 API Availability
OpenAI's language model, ChatGPT, has experienced some turbulence in its user growth. Data from SimilarWeb shows a decrease in the number of users from May to June, potentially indicating a stagnation or even a decrease in user engagement. In the midst of these changes, OpenAI has worked hard to ensure that most people who signed up for API access can now leverage the power of GPT4 using tools that can access the API (e.g. Zapier).
AI-Generated Art Protection and Initiatives
There have been some recent initiatives focused on protecting artists' work from being used without permission by AI art generators. C2PA uses cryptography to tag AI-generated content, while Photoguard tweaks image pixels to prevent AI models like Stable Diffusion from recognising them. These are interesting tools, but still unproven in their ability to sidestep AI systems built to create art. As marketers create more visual content with AI, being informed on both the copyright issues and the need to protect your own original art and creative work is increasingly important.
Meta's AI "Personas" and Digital Marketing
Meta has developed AI-powered chatbots with unique personas for platforms like Facebook and Instagram. These bots aim to boost engagement through personalized recommendations and conversational features. For marketers, they present opportunities to better understand target audiences and deliver tailored experiences. However, increased data gathering and use for ad targeting raises privacy concerns. Marketers should watch Meta's approach to ensure it respects user consent and boundaries, while also looking for opportunities to better understand, serve and engage with their target audiences.
Instagram's AI-Powered Features
Instagram is reportedly developing new AI features like image labeling, DM summarization, and enhanced editing tools. Labeling could help curb misinformation and establish trust around real versus AI-generated images. DM summaries would help brands and influencers manage conversations at scale. Editing advancements may enable easier high-quality content creation. As AI plays a bigger role across social platforms, understanding emerging capabilities on Instagram can help marketers supercharge their approach.
Google Assistant's AI Upgrade
Google is planning a major upgrade to its Google Assistant by incorporating generative AI, according to an internal email. This would give Assistant similar capabilities as ChatGPT and Bard. With Assistant's wide consumer reach, this revamp makes the platform increasingly important for marketers as a channel for engaging people through conversational interactions.
OpenAI's GPT-5 Trademark and Its Implications
OpenAI has filed a trademark application for GPT-5, which, among other things, covers various speech processing capabilities. This points to efforts to evolve ChatGPT from text to voice interactions, more akin to how people interact with computers in futuristic TV shows like Star Trek (i.e. by just talking to them). However, OpenAI's roadmap and timing for actually releasing GPT-5 remains unclear.
Meta Open-Sources Audiocraft for AI Sound Creation
Meta open-sourced AudioCraft, a text-to-audio tool that it created for generating music, sound effects, and more. This opens up possibilities for marketers to create a wide range of audio for different use cases (royalty free). AudioCraft is not yet available for marketers (or anyone else in the wider market) to tinker with, but watch this space!
Interview with GoCharlie's Brennan Woodruff
At the end of this episode, Brennan Woodruff, Co-Founder and Chief Business Officer at GoCharlie, discussed the startup's AI content engine for marketing teams. He outlined challenges in building an AI company amid intense hype and competition. Woodruff shared examples of workflows combining AI models for powerful results. He also explored the future potential of AI agents running autonomous tasks proactively. Overall, Woodruff explained GoCharlie's mission to empower businesses and people through AI automation in marketing content creation and beyond.
Summary of episode 21:
Unleashing the Power of Spoken Emotion with PlayHT2.0
PlayHT has released a major update to its conversational AI voice model called PlayHT 2.0. It can mimic voices and accents with high realism after training on just a few seconds of sample audio. The model also allows users to control emotional expression, such as happiness or fear, when synthesizing speech. This promises huge potential for applications like virtual assistants and content creation (e.g. human-sounding audiobooks, voiceovers etc.).
AI Avatars Go Viral with Lifelike Realism
Startup HyperHuman has unveiled HyperHuman 2.0, its new AI avatar clone model. The system can create incredibly lifelike avatar videos of someone after training on just two minutes of sample footage. The results are so realistic that people are questioning whether the avatars are actual humans. This points to an explosion of synthetic video content across media formats and is definitely a trend for marketers to keep an eye on.
ChatSpot's AI Upgrade: Summarizing CRM Records and More
HubSpot's conversational AI tool ChatSpot has been upgraded with new features like summarizing CRM records and YouTube videos, as well as a number of other improvements. The progress being made suggests that ChatSpot is reaching a tipping point of true utility for marketers (rather than the interesting but limited alpha release it has been so far).
Stability AI's New Beluga Models Set Benchmark Records
Stability AI released two powerful new AI language models called Stable Beluga 1 and 2. They were trained on less data than comparable models, but benchmark very highly on assessments of language understanding and reasoning. More competition and advancement in open source models provides commercial options for chatbots and other AI applications.
Anthropic's New Model Could Help Marketers Generate Better Content
Anthropic has updated its Claude Instant model to version 1.2, incorporating strengths from Claude 2 such as better math, reasoning, and longer text generation (including a 100K context window). For marketers, smaller faster models that keep improving will enable more cost-effective content creation APIs.
Longshot.ai Launches BotShot for Training Bots on Company Data
AI company Longshot has launched BotShot, a no-code solution for creating chatbots trained on proprietary company data. This makes it easy for any business to leverage the power of large language models tailored to their organization's data and internal information without technical skills.
Zoom Updates Terms to Train AI with Customer Data
Zoom has updated its terms of service to allow using customer data like product usage telemetry to train its AI models. However, it states audio, video and chat content won't be used without consent. As AI needs vast data, companies are tapping user data, but privacy concerns remain.
AgentBench Evaluates LLMs as Software Agents
Researchers have developed AgentBench, a new benchmark to evaluate large language models at completing real world tasks across 8 environments. It tested 25 models with OpenAI's GPT-4 scoring the highest, showing its greater general intelligence. So it looks like GPT-4 is still likely to be the model of choice for most marketers requiring high-quality outputs.
MetaGPT Aims to Replace Entire Software Companies
MetaGPT coordinates different AI models in specialized roles like project manager and engineer to collaboratively build software driven by short, simple user prompts. In one example, it created a Flappy Bird game with just a one sentence prompt. As agent coordination advances, creating apps and digital projects becomes exponentially easier. As marketers, we may quickly be moving to a world where we are no longer restricted by the tools available to us... instead, we may be able to collaborate with an AI bot to build the software tool or plugin we need to serve our unique needs.
Summary of episode 22:
Generative AI Reaches Peak Hype: What Marketers Need to Know
According to Gartner's 2023 Hype Cycle, generative AI like ChatGPT has reached peak hype following $14 billion in investment and 100 million monthly active ChatGPT users. However, uncertainties around security, copyright, "hallucinations", and impending regulation could lead to a "trough of disillusionment." Marketers should avoid getting overly caught up in the hype and pragmatically integrate AI into workflows rather than fearing AI will replace them.
Our feeling is that some aspect of AI have matured enough to have already crossed the "trough", and that lumping all generative AI tools together misses importance nuances about what is already ready for commercial use (and what is not).
McKinsey's Chatbot Lilli: The AI-Driven Consulting Tool
McKinsey has developed an internal AI chatbot named Lilli, trained on over 100,000 pages of their internal knowledge base to provide expert advice for internal McKinsey team members to utilise by pulling research, case studies, and more. This demonstrates the value of centralizing and documenting company knowledge into easily accessible formats. Such chatbots can enhance employee efficiency, accuracy, insight and creativity.
Google's Warning: AI-Generated Content Often Rehashed Material
Google's John Mueller warned SEO pros that AI-generated content is inherently rehashed from other sites. Google wants fresh content that adds to the body of knowledge (not rehashed existing info and insights). While AI-generated content can help with ranking, it often fails to meet Google's original research and insight criteria. Therefore, it is essential for marketers to prioritize the creation of captivating and exceptional content that provides unique value and insights to searchers and readers.
Microsoft Launches Azure ChatGPT: A Secure, Private Alternative for Enterprises
Microsoft unveiled Azure ChatGPT, an enterprise version of ChatGPT hosted privately on Microsoft servers. This provides a secure alternative for businesses hesitant to use public AI models using their proprietary information and data. The move signals tech giants scrambling to corner the enterprise AI market and provide larger businesses with a secure chatbot option.
Google's Smarter AI Features Revolutionize Search Experience
Google Search's new generative AI features can summarize webpages and key points while users browse. This aims to expedite researching complex topics, but may also reduce organic traffic to websites, as answers appear without webpage links. Marketers relying on SEO must adapt to AI-driven changes in how Google surfaces and highlights content.
Amazon's AI Revolutionizes Product Review
Amazon is using AI to create highlights from customer reviews, providing quick product sentiment snapshots from verified purchases. This makes sense given Amazon's reliance on reviews for purchase decisions. It further shows the importance of cultivating reviews, as AI summarizes their sentiment and content for users.
AI can be more creative than humans
Studies found GPT-4 surpasses most humans at creativity tests, also generating more feasible ideas. Yet top creative individuals can still outperform it. This makes AI a promising creative brainstorming tool rather than just a content production tool. Marketers can leverage it as an ideation collaborator, especially if creativity isn't their strength.
Interview with Robert Rose: Content Marketing Institute Career Outlook Report 2024
Robert Rose, Chief Strategy Advisor at the Content Marketing Institute, discussed the company's 2024 Career Outlook Report focusing on AI's impact on content marketing roles. He outlined mixed feelings among marketers on using AI writing tools like ChatGPT, despite appreciating their value. Rose explained concerns around potential job loss and lower compensation. He recommended differentiating yourself through strategic expertise, business acumen, and editorial judgment (not just writing skills). Rose also called for companies to provide better career advancement paths for content marketers. Overall, he provided perspective on both optimism and challenges regarding AI's emergence in the content marketing field.
Summary of episode 23:
Google's TextFX: A Creative Writing Toolbox
Google has launched a new AI tool called TextFX designed to assist creative writers. It includes features like "Unfold", which inserts words into phrases, "Similes" for generating metaphors, "Explode" to break down words into similar sounds, and "Chain" to build semantic connections. TextFX aims to provide writers with AI-generated ideas and angles to spark creativity. However, it can sometimes provide irrelevant or unhelpful suggestions, so human oversight remains critical. Overall, TextFX offers writers a toolbox of AI features to experiment with and integrate as part of their creative process.
AI Art Denied Copyright Protection
A recent US court ruling denied copyright protection for art created solely by AI systems that did not have clear human input. The verdict referenced legal precedents requiring human creativity in copyrighted works. As AI-generated art proliferates, this raises questions around legal ownership and protections. While AI can be used as part of a creative process, works produced entirely by AI may not qualify for copyright. This indicates creators should carefully document human contributions when using AI tools to potentially secure copyright protections.
GoDaddy's AI-Generated Videos for Entrepreneurs
Web hosting company GoDaddy launched GoDaddy Studio to help entrepreneurs easily create marketing videos using AI and templates. Features include background removal, instant video creation from photos/clips, and AI-generated text overlays. While somewhat basic in terms of functionality, it provides an accessible entry point for small business owners to make video content for social media. For sophisticated video production, other tools like Descript, Wideo or Kapwing offer more advanced creative options. But GoDaddy Studio makes video creation achievable for non-expert users.
AI Skills on the Rise, But People Skills Prevail
AI terms like "ChatGPT" and "prompt engineer" have surged in popularity on LinkedIn profiles, indicating rising demand for AI talents. However, LinkedIn research found interpersonal skills remain the most valued by executives. Flexibility, ethics, social awareness and self-management topped the list of sought-after capabilities. This shows that while AI expertise has cachet, uniquely human strengths like communication, reliability and strategic thinking remain indispensable. AI augments but doesn't replace the multifaceted capabilities marketers bring through their expertise and unique experience.
Adapting to AI: The Workplace Evolution
An IBM study of global executives and employees suggests AI won't replace human roles, but that many people will require reskilling. Forty percent of the workforce may need to learn new skills within three years as AI's footprint expands. But most leaders view AI as enhancing versus supplanting jobs. When hiring, soft skills like teamwork and communication are likely to be prioritized over technological capabilities. This indicates successfully leveraging AI relies on uniquely human strengths like creativity and empathy. Those able to strategically apply AI while cultivating emotional intelligence will lead the future workforce.
SeamlessM4T: Bridging Communication with Multimodal Translation
Meta announced SeamlessM4T, an AI system for translating between languages via speech, text and speech-to-text. It supports 100 languages with minimal lag between input and translated output. This technology has exciting implications for marketing communications and campaigns. Brands can rapidly localize content and connect cross-culturally. However, it raises concerns about how language learning may evolve given AI's translation capabilities. While promising for business efficiency, a general translator could reduce incentives to learn languages.
AI Content Detectors and Limitations
Studies show current AI detectors often fail to reliably identify AI-generated text, while frequently mislabeling human writing. Detectors also exhibit bias against non-native English patterns. This suggests Google is unlikely to deploy unreliable detectors that would incorrectly flag quality content. For SEO, focus should remain on creating original, valuable content versus trying to "trick" detectors. AI can assist writing, but human creativity, strategy and editing ensure content quality and value to readers.
AI Art with Midjourney's New Inpainting Feature
Midjourney, the AI art generator, introduced a feature called Inpainting for editing images. Users can select a portion of an AI-generated image and rewrite the prompt to modify just that section. This brings more flexibility compared to re-generating entire images from scratch. For designers and marketers, Inpainting enables greater control over Midjourney's creative outputs. Production workflows with AI tools become more efficient by updating specific parts rather than starting over.
Fine-Tuning ChatGPT: A Business Game-Changer?
OpenAI has released fine-tuning capabilities for ChatGPT to train it on specific company data and language patterns. This allows businesses to customize ChatGPT models for unique use cases like customer service chatbots, which are trained on the company's own data. However, fine-tuning requires high-quality structured data and technical expertise. While promising for advanced AI applications, the average marketer likely lacks the data science skills to fine-tune ChatGPT independently. But outsourcing fine-tuning could enable impactful business use cases.
AI Revolution in Advertising
Major brands are actively experimenting with generative AI for advertising. Campaign examples include AI-generated video personalization and using AI to craft product descriptions at scale. WPP and Nestle reported positive results from testing AI content creation. While cost and efficiency gains are appealing, the most successful early uses have empower audiences to co-create tailored assets, leveraged AI to make localised, interactive campaigns possible for businesses of all sizes (and making the campaigns go viral in the process).
AI Bots Surpass Humans in CAPTCHA Tests
Research from UC Berkeley and Microsoft indicates AI bots can now solve CAPTCHA tests with up to 100% accuracy in about 1 second. This surpasses the 50-85% human accuracy rate of around 15 seconds per CAPTCHA. Even Google's more advanced reCAPTCHA is solved by bots with 85% accuracy in 17 seconds, on par with the 18 seconds for humans. For marketers, as AI progress renders CAPTCHA ineffective, alternative bot detection options will be needed to secure online forms and prevent spam.
Interview with Louis Pereira, Founder of Audiopen
Louis Pereira, founder of the AI-powered voice transcription app AudioPen, discussed his journey creating the tool. He explained Audiopen's capabilities for effortless note-taking and content creation through voice. Pereira highlighted his focus on an intuitive, user-friendly interface and shared how he leveraged no-code solutions like Bubble to build and rapidly iterate Audiopen. Pereira also provided perspective on AI's promise and risks, emphasizing the need for human creativity and problem-solving. He encouraged experimentation with AI to uncover impactful solutions.
Explore further with the "Artificially Intelligent Marketing" Podcast
This blog provides just a snapshot of the invaluable insights Paul and Martin discuss in their engaging podcast episodes. For an in-depth understanding of the fascinating realm of AI in marketing, we highly recommend listening to the full episodes, either here or on your podcast platform of choice. Additionally, the archive of past episodes is a deep pool of expertise on how AI is reshaping the marketing domain.
The development of this blog post was facilitated by AI technologies. Initially, the podcast audio was transcribed through the use of AI, then Anthropic's Claude and GPT-4 (ChatGPT) aided in condensing the transcript into concise blog sections. Our BioStrata team then carefully revised and fine-tuned these segments for maximum clarity and flow. Furthermore, AI software from Clipdrop was instrumental in generating the blog's accompanying image.