This is a continuation of my AI primer for marketing professionals series. The first part looks at ChatGPT, Dall-E 2, Codex and more on OpenAIs current and iterative AI models. Be sure to check out Part 1: What is OpenAI, ChatGPT, Dall-E 2 and How Do You Use Them
Open AI Creates a DotA2 player that shreds pros
In 2017 OpenAI developed a ‘DotA 2 Player’ that was designed to learn and develop competitive abilities. The result is that the ‘bot’ began to develop truly expert techniques and competitive ability.
Teams that played the bot were so enthused, they sought to integrate the bot in their regular training and improvement practices. The implications this has for devising automated solutions in messy, complicated real-time human situations is incredible long-term.
The AIOpen DotA’ player, “Dendi” debuted at The International and went undefeated against T1 teams including, SumaiL and Arteezy, both with Evil Geniuses and top world ranked players at the time.
What are modal ‘neural networks’ Like CLIP – AI training
CLIP is OpenAI’s ‘neural network’ that learns visual concepts from natural language inputs. Through these methods, AI can learn from visual inputs, images and other visual resources to define its learning. Stick with me, this gets a bit nerdy.
Clip stands for Contrastive Language-Image Pre-training and seems to be a persistent work in progress. The simplest way I could understand it is the process of getting AI to learn on its own through derivative logic and rational based on what it learns about what it sees.
On this path, the term ‘Zero-shot’ comes up a lot.
Zero-shot is a machine learning classification which is, in short, a machine’s ability to rationalize and learn a concept without any previous specific ‘training.’ AI uses previously seen examples and training to produce a logical understanding of an unseen or previously unknown problem, input, query etc.
The process relies on enormous visual and text databases like ImageNet and WordNet that matches words, nouns and syntax with millions of images and descriptive visual aids forming the basic ‘learning repertoire’ of AI models.
The process creates and relies on “multimodal neurons” – a type of neural network in visual and natural language processing to analyze and comprehend information from multiple stimuli like sound, look and text or copy.
This allows a model to describe an image it sees or produce a visual representation of a description and much, much more.
I am aware that this sounds like it was written by a robot but it was me, Brad, the human. I only understand this part from my experience as an SEO, diving deeper into natural language processing as it relates to Google search and AI. The principles are very similar.
What marketing professions are best suited to use and integrate AI
Of course, ALL professionals should have some use for AI tools, but as I went through this experience there are some marketing practices that I felt are very naturally equipped to conduct AI integration and practice.
SEOs and search professionals:
SEOs are honed to language trends that people use when they want specific information. This deep understanding and commitment to analyzing how to ask the right questions and yield the right results is a perfect pairing to getting AI models to get, give and produce results you want.
Creative and copywriting professionals:
Let’s just agree to be real about this. ChatGPT is not the T-1000 of your profession. That won’t stop orgs like Buzzfeed from making stupid decisions around it. Very quickly you, I and all of the content consuming public will be able to currently recognize something that ChatGPT spit out compared to something a writer with decades of experience crafted.
However, it can be a valuable tool to eliminate about 70% of the heavy lifting that comes with crafting written materials. OpenAI is also actively designing class attributions to detect dishonest uses of ChatGPT to help combat abusive use in academics and professional/protected settings.
Web and desktop site publishers:
Codex will continue to grow in scope and power. While this will always amplify the power of a skilled coder and designer, it can empower desktop and web publishers to likely get through moderate coding issues without having to add to your design team’s tickets and workload.
That said, this will have impact on the staffing of code teams themselves.
Reporting, Analytics and BI teams:
If you are in a discipline where you love preparing reports and data insights but hate or struggle with the writing part. ChatGPT is a godsend. You can focus on generating the reports and compiling the data and get ChatGPT to add the summary explanations or context.
**In all cases, I’ll reiterate that you should avoid actually submitting sensitive or private data. What goes in ChatGPT does NOT stay in ChatGPT**
What’s new and coming up for ChatGPT and OpenAI
As the OpenAI team (and likely everyone else now paying attention) hurtles toward rapid advances in AI technology, here are some of the upcoming milestone items we can expect to see in the near future.
Cost model for ChatGPT and like models:
There is already the framework for the ChatGPT Plus subscription plan, which will look to be about $20 USD per month. It will allow general access to ChatGPT during peak times, faster response times and priority access to new features and improvements.
ChatGPT Plus is currently only available in the US but OpenAI is working on rolling it out to their waitlist in other countries and regions. Free users will not be affected by this cost model.
AI generated work classification and detection:
OpenAI is working on an AI Classification to determine and identify AI-created text and products to combat academic dishonesty and combat misuse or misinformation.
It begs the question about what implications this could have for professional use? Would clients hesitate to pay rates for AI generated copy? Do clients have a right to know if and when AI is used to produce work on their accounts?
What about in the workplace? Is your boss entitled to know if you used ChatGPT to complete a presentation, pitch or assignment?
This classifier is an important step towards transparency and authenticity in our work and the role of AI in it.
OpenAI Whisper – text transcription and advanced voice-to-text (VTT):
Whisper is OpenAI’s automatic speech recognition system or ASR. Nearly a million hours of multi-lingual data is being used to train transcription against accents, speech nuances, background noise and voice isolation and more.
Whisper even supports transcription in multiple languages and translation of those languages into English. To-date, 1/3 of the dataset for Whisper is non-English, so it’s not going to overlook non-English speaking users, projects and requirements.
Whisper was introduced in September 2022 and is currently in beta, not released for public trial or research preview.
OpenAI’s mission is to ensure that artificial intelligence benefits all of humanity. An important part of this effort is training AI systems to do what humans want. As mentioned, the group has an aggressive mandate to combat misuse, harmful content and misinformation.
I know these can be scary times, but I also feel it’s important to embrace the steady march of tech advances and the implications it has on our working world and life. It can be a great tool with the right application and understanding.
I understand the hard parts. I am a digital marketing professional. AI and marketing automation has persistently taken over the work I would have traditionally done on a near monthly basis. I’ve learned that the best asset I can provide is the willingness to understand, master and implement these changes, not rage against them.
It may sound harsh, but my belief is that there will emerge two camps in the professional world as AI continues its lightspeed development and integration:
- Those who rage, fight against and resist AI in the interest of pushing it back, unsuccessfully and
- Those who embrace AI as a tool to amplify and integrate with modern service requirements in an effort to master implementation and generate desired results
I suppose it’s obvious what camp I’m in.