by Reuben Maher
ASHBURN, VA, February 28, 2023 — Backstage in the brightly lit green room where my band and I were waiting to head onstage for our next set, the conversation had pivoted from music and movies to technology, and whether or not Artificial Intelligence (AI) and Machine Learning (ML) were slowly taking over the earth. The focus of our attention was the recently-released ChatGPT – a program that uses AI to understand and generate human-like language responses – and how this and other new AI capabilities were impacting not only the world around us, but also our livelihoods. One of the players talked about his role as a subject matter expert in the graphic design and branding industry. He was concerned that AI will now replace the value he and his colleagues bring – with decades of experience – due to AI’s ability to rapidly generate realistic art and imagery at very high levels of quality and creativity. Another band member hadn’t heard of ChatGPT and wondered if it could help her generate a new tagline for her business – which it successfully did over the next two minutes based on only a couple of tailored prompts. Another person expressed excitement about the possibility of leveraging ChatGPT as a personal research assistant in his job. Overall, the feelings varied from deep suspicion and grave concern to enthusiastic optimism about the art of the possible.
ChatGPT is rapidly gaining notoriety and becoming more mainstream in overall usage. Notably, ChatGPT is currently outpacing TikTok in its growth, as it surpassed 100 million users in January 2023 after only 2 months – a milestone that took TikTok nine months and Instagram two and a half years to reach.
Part of the reason for ChatGPT’s success is not only the advances in large language models (LLMs), but also in the additional optimization the model undergoes through the Reinforcement Learning from Human Feedback (RLHF) method. Through the RLHF process, the model gets further refined to sound much more human-like in its conversation through a collaborative training process which engages a large number of human overseers to test and refine it.
Such human-machine training is a key enabler for how powerfully realistic ChatGPT’s conversational AI interactivity has become and is thus the driving factor behind its meteoric rise in popularity. You can ask it virtually anything from the mundane to profound… personal to professional… funny to serious… and be amazed at what it returns. Interesting prompts include:
Translate the following text into [pick your language]
As a chef, give me recipes for every night this week that I can cook for my family that are healthy
Write a love poem to my spouse named [name] in the genre of an Eminem rap
Pretend you work in the accounts receivable department, and you need to collect payment from a client that owes you $50,000. Write a scary email that encourages them to make prompt payment in the next 30 days.
Create a funny sketch between Spider-Man and Superman having them talk about who is the better superhero between the two of them
How can I grow my company’s audience on social media [and for more specific advice, pick the social media platform(s)]?
Act as my tutor. Help me understand how Taylor’s theorem in calculus works. Describe it in easy-to-understand terms.
Not to be outdone, many similar conversational AI projects and research results are being announced and quickly released into the marketplace. In early February, Google revealed the imminent widespread release of its Bard chatbot, powered by its Language Model for Dialogue Applications (LaMDA) ML language technology. After Microsoft – an early and ongoing investor in OpenAI, the AI research laboratory that created ChatGPT and was founded by tech titans such as Elon Musk, Reid Hoffman, Peter Thiel, and Sam Altman – announced its Bing app was integrating ChatGPT into its search capability, the Bing app’s audience grew nearly as much in the following week (750,000 downloads) as it had the entire previous year (800,000 downloads). In its 2023 MAD (Machine Learning, Artificial Intelligence and Data) Landscape report released in late February, Firstmark detailed “the unmistakable, exponential acceleration of Generative AI…. [and its] deep impact on society, politics, geopolitics and ethics.”
In addition to conversational AI, advanced AI/ML technology is being applied to the arts, producing stunning unique images and audio/musical compositions. OpenAI’s DALL-E is the most well-known of the AI image generators as it “can create original, realistic images and art from a text description.” If you’re like me, you may be skeptical. Yet the image at the top of this article was generated by DALL-E in response to a focused text prompt I entered. Additionally, in the last week of January, four different audio models were released, including MusicLM and MusicCaps (by GoogleAI), Moûsai, and AudioLDM, which generate original audio and/or music based on text-guided instructions from the user.
As incredible and, to some, overwhelming as the increasingly rapid evolution of AI is, responses from any AI platform need to be vetted for accuracy as they can be imperfect…and, the impacts of the occasional incorrect response can be significant. In early February of this year, an ad for Google’s new Bard capability contained an error that prompted a nearly 8% selloff in Google shares, obliterating more than $100 billion in market value! There are numerous reports of kids trying to get ahead using ChatGPT to complete their homework only to have it return answers that are nonsensical or incorrect.
As incredible and, to some, overwhelming as the increasingly rapid evolution of AI is, responses from any AI platform need to be vetted for accuracy as they can be imperfect…and, the impacts of the occasional incorrect response can be significant.
However, with all of the challenges, we’re still only seeing the tip of the iceberg regarding AI’s conversational capabilities. ChatGPT currently has 175 billion parameters. Sound impressive? The next generation called GPT-4 is estimated to have 100 trillion parameters! (For context, the more parameters the model contains, the faster it should be able to generate more accurate and relevant responses.)
In that vein, I was curious to find out how accurate ChatGPT would be in analyzing its competition and entered this simple prompt:
“Write a 500 word blog post about the differences between ChatGPT and the new Google Bard capability.”
I’ve included its response below in its entirety with no editorial touch-ups. But feel free to verify my assertion. Simply visit GPTZero – an app written by a 22-year-old Princeton University senior to detect whether or not text has been written by AI.
The irony is not lost on me – AI checking on other AI. What could possibly go wrong?
*******RESPONSE GENERATED BY CHATGPT*******
Recently, OpenAI introduced its new language model, ChatGPT, which has been making waves in the field of artificial intelligence and natural language processing. On the other hand, Google has also introduced a new language model called Google Bard, which promises to revolutionize the way we interact with technology. In this article, we will compare and contrast the two models to give you a better understanding of the differences between them.
First of all, let's take a look at the architecture of the two models. ChatGPT is a transformer-based language model that uses a deep neural network to process and generate text. The model has been trained on a massive amount of data, allowing it to generate high-quality text that is often indistinguishable from human writing. On the other hand, Google Bard is a smaller, lighter language model that has been designed specifically for use in conversational interfaces such as chatbots and voice assistants. Unlike ChatGPT, which is designed for a wide range of tasks, Google Bard has been optimized for a specific set of use cases.
One of the main differences between the two models is the amount of computational resources required to run them. ChatGPT is a very large model, and running it requires significant computational resources. This makes it less suitable for use in real-time applications, such as chatbots and voice assistants, where response times need to be fast. In contrast, Google Bard is a much smaller model that can be run on much less powerful hardware, making it much more suitable for real-time applications.
Another difference between the two models is the quality of the text they generate. As mentioned earlier, ChatGPT has been trained on a massive amount of data, and it generates text that is often of high quality. This makes it ideal for use in applications such as content creation, where high-quality text is a must. On the other hand, Google Bard is designed specifically for use in conversational interfaces, and it generates text that is optimized for readability and conversational flow. While the quality of the text generated by Google Bard may not be as high as that generated by ChatGPT, it is still of high enough quality to be used in real-world applications.
In terms of speed and responsiveness, Google Bard has a clear advantage over ChatGPT. Because it is a smaller model, it can generate responses much faster than ChatGPT. This makes it ideal for use in real-time applications such as chatbots and voice assistants, where fast response times are essential. In contrast, ChatGPT is much slower, and it may take several seconds to generate a response.
Finally, let's talk about the cost of using the two models. ChatGPT is a complex model that requires a significant amount of computational resources to run. This means that it is relatively expensive to use, and it may not be suitable for use in applications where cost is a concern. On the other hand, Google Bard is a smaller, simpler model that can be run on much less powerful hardware. This makes it much more cost-effective to use, and it is a good choice for applications where cost is a concern.
In conclusion, ChatGPT and Google Bard are two very different models, each with its own strengths and weaknesses. ChatGPT is a large, complex model that generates high-quality text, but it is expensive to run and may not be suitable for real-time applications. Google Bard is a smaller, simpler model that is optimized for use in conversational interfaces, and it is much more cost-effective and suitable for real-time applications. Ultimately, the choice between the two models will depend on the specific needs of your application, and it is important to carefully consider the trade-offs between quality, speed, and cost when making a decision.
In terms of applications, ChatGPT is well suited for content creation, while Google Bard is better suited for conversational interfaces. Both models are powerful tools that have the potential to change the way we interact with technology, and they represent exciting developments in the field of artificial intelligence and natural language processing. Whether you are looking to create high-quality content, or you want to build a chatbot or voice assistant, ChatGPT and Google Bard are both worth considering.
In conclusion, ChatGPT and Google Bard are two cutting-edge language models that each offer unique advantages and disadvantages. Understanding the differences between them will help you make informed decisions about which one is best suited for your specific needs and requirements. Whether you are a researcher, developer, or business owner, keeping an eye on these models and their continued development is a must for anyone interested in the future of artificial intelligence and natural language processing.
*****************************************************
About Skybrid Solutions
Skybrid Solutions is a minority-owned and SBA-certified Woman-Owned Small Business (WOSB) serving both public and private sector organizations. Skybrid is focused on delivering digital modernization capabilities including full stack and agile development, analytic intelligence, cybersecurity, and cloud transformation. The release of Skybrid Flare allows distributed teams to rapidly deploy an automated DevSecOps innovation platform leveraging a zero trust model. The Skybrid team brings an extensive track record of harnessing emerging technologies to create transformative solutions and extraordinary customer experiences for its clients. Visit skybrid.com and connect with Skybrid on LinkedIn, Twitter, Instagram, and Facebook. Additional information is also available in Skybrid’s unique 3D (virtual reality (VR)) / 2D (web browser) metaverse environment – Skybrid Island – which transforms its web presence into a VR experience offering a glimpse into Skybrid’s forward-leaning innovation and best-in-class technology capabilities.
Contact Information:
Skybrid Solutions, LLC
Comentarios