The Present & The Future of AI-Generated Videos – a Discussion with Synthesia

22

October

2023

No ratings yet.

Imagine a world where you can create captivating, high-quality video content at the speed of thought. AI-generated videos are now transforming this vision into a reality. These digital platforms can turn basic scripts – and generate the scripts themselves, as well – into stunning masterpieces, adapting to various languages and global audiences through diverse avatars and instant translation options. 

To delve into this cutting-edge technology, I chose to test out Synthesia, a leading AI video platform, and a unicorn with a $1 billion valuation (UCL, 2023). With features like automatic script-to-video conversion, multilingual support, and customization (Synthesia – AI Video Generation Platform, n.d.) as well as an impeccable customer list, I was intrigued.

I delved deep into researching the company and the ethics of AI-generated content, generated two demo videos, and had an interview with Borys Khodan, the Head of Paid Media at Synthesia, for a comprehensive research about the company’s direction and ethical considerations.

The Synthesia Experience

To start with, the overall customer experience for demo videos at the moment seems heavily targeted towards B2B. The clients list of Synthesia features well-known companies, including Amazon, Zoom, BBC etc., and not individuals. Organizations collaborating with Synthesia often use it to create how-to videos, used for learning & training, IT, sales, and marketing teams. There are a ton of features available for them. What truly caught my attention is the effortless video updates – if a script of a video has to be changed, a simple script tweak yields a new video almost instantly, sparing companies from the costly ordeal of re-recording content from scratch. The potential savings here are impressive and certainly make a difference in the corporate landscape. Moreover, Synthesia has plans to expand its offerings for individuals (B2C), introducing advanced 3D avatars and other features for additional use cases, as confirmed by Borys.

Source: Synthesia.io

My personal experience involved choosing one of Synthesia’s demo video options, “Sales Pitch.” I edited the standard text and received the video via email within minutes. It was an impressive process, though there was a minor glitch. While the AI handled the standardized text perfectly, it struggled to pick up the word “Resume” as a noun and mispronounced it as the verb, when I attempted to generate a sample pitch for a potential employer. However, in the paid version, there must be tools to easily adjust the videos. When I tried the “compliment” option to generate another video, it worked smoothly – sending a personalized video to me directly without any hiccups. It was quite fun to watch the text that I wrote in 15 seconds come to life in a video! 

AI-Generated Video: Sales Pitch

Industry Overview

In this rapidly evolving landscape, competition among AI-driven video generation platforms is fierce, leading to heightened ethical concerns. Lots of companies scape the web for data – sometimes, using personal data for training their algorithms to get ahead of the competition. But for Synthesia, the path is clear – no shortcuts, no compromises. The company’s founders have committed substantial resources to ensure ethical data procurement, even if it may mean encountering short-term delays and additional expenses. The company wants to be sure that the data they use to train their models on are ethical, legal, and are not violating any privacy laws (Browne, 2023). This approach may shield Synthesia from potential legal repercussions stemming from unethical or illegal data sources.

Ethical Implications

This ethical lens extends to the creation of AI-generated videos, and brings us to the questions of authorship, artistic authenticity, and ownership. What if AI is requested to generate violent or inappropriate content? What happens when an AI can replicate the voice and likeness of a person without their consent? The concept of deepfakes, where AI manipulates content to deceive and misinform, also poses a clear danger, and there are many opportunities and threats in that direction (Collins, 2023). I asked some of these questions to Borys in a fruitful discussion about the future of AI-generated videos. 

“We maintain very high ethical standards and closely monitor all content generated by our customers. If the content violates any of our policies, such as featuring violence, harmful behavior, offensive language and more, the video would not be generated,” Borys said. “Our high ethical standards aren’t just nice to have; they are core to our mission and set us apart in a competitive market.”

Unfortunately, not everyone in the industry follows these standards, making the next few years crucial for determining who will succeed, and what legal requirements will evolve. For now, there are no current regulations requiring content to be labeled as AI-generated, but that may change in the future.

Conclusion

AI-generated videos hold immense potential for reshaping the corporate landscape. Synthesia serves as an inspiring example of this transformation, and working even with their demo product was seamless, quick and simple. However, harnessing the potential of the technology to the fullest extent requires unwavering adherence to strict ethical standards of all the companies in the sector, reminiscent of Synthesia’s own commitment. When wielded responsibly, AI-generated videos stand as a remarkable testament to the synergy of technology and business solutions, marking an important step into the future.

AI-Generated Video: Compliment

References

Browne, R. (2023, June 13). Nvidia-backed platform that turns text into A.I.-generated avatars boosts valuation to $1 billion. CNBC. https://www.cnbc.com/2023/06/13/ai-firm-synthesia-hits-1-billion-valuation-in-nvidia-backed-series-c.html

Collins, T. (2023, April 24). The Rise of Ethical Concerns about AI Content Creation: A Call to Action. IEEE Computer Society. https://www.computer.org/publications/tech-news/trends/ethical-concerns-on-ai-content-creation

Synthesia – AI Video generation platform. (n.d.). Www.synthesia.io. https://www.synthesia.io/

UCL. (2023, June 15). AI firm Synthesia, co-founded by UCL scientist, becomes $1bn unicorn. UCL News. https://www.ucl.ac.uk/news/2023/jun/ai-firm-synthesia-co-founded-ucl-scientist-becomes-1bn-unicorn

Please rate this

Do we need to learn languages anymore?

20

October

2023

No ratings yet.

My previous post talked about my experiences with using Generative AI tools for learning languages. I received a comment on the blog which talked about if there will even be a need to learn languages when advanced technology can eventually translate any language with near perfection in real-time into your ears. This image attached is an AI’s imagination of people who know no common languages communicating with the help of an instant translator earpiece. This is a daunting proposition, that there will be fewer and fewer language diversity as times go on due to the domination of the internet by a few languages, mostly English.

Generative AI is a lot different from a mere translator application as it can understand the nuances of a language and translate in a much more accurate way. But as more of these tools are majorly trained in English, the lingua franca of the world, there is a possibility of widening gap between English and some European languages vs the rest of the languages (Sharma, 2023). ChatGPT for example could reinforce the dominance of English, and result in reduced richness of language as it does not currently function well with non-European languages.

There is hope though as it has been observed that Generative AI apps have been able to comprehend languages that they were never initially trained on. The process of how AI is able to pick up languages is not explained currently however there are attempts to explain it using the way humans pick up languages. Awareness of a single languages predisposes humans to learn a second language merely through immersion in the second language, as we’re able to recognize patterns in the language to understand its structure. The same has been considered in the case of Generative AI which is able to recognize patterns in languages as well, suggesting that as AI learns more languages, it may be easier for it to catch on even more (Eliot, 2023). However, the resources available for certain languages far exceed that of others and the risk of losing linguistic diversity is still exemplified in the age of Generative AI due to this difference.

As a language enthusiast myself, I see myself reliant on Google lens to translate even the smallest of foreign language words when I am in a new country, which removes the need to even read or type it into the translator, thereby reducing all the interaction I could have with the new language. Do you have any experience where you find yourself using lesser linguistic power as Generative AI can do it for you?

References:

Sharma, Y. (2023, February 24). What CHATGPT means for linguistic diversity and language learning. University World News. https://www.universityworldnews.com/post.php?story=20230223132734294

Eliot, L. (2023, April 19). Solving the mystery of how chatgpt and Generative AI can surprisingly pick up foreign languages, says AI ethics and AI Law. Forbes. https://www.forbes.com/sites/lanceeliot/2023/04/19/solving-the-mystery-of-how-chatgpt-and-generative-ai-can-surprisingly-pick-up-foreign-languages-says-ai-ethics-and-ai-law/?sh=5aab7cc36b1d

Please rate this

ChatGPT Plugins: The Doom or The Boom?

17

October

2023

5/5 (2)

Please rate this

The Promise of Plugins

The emergence of ChatGPT plugins has been a turning point in the realm of conversational AI. Not only do plugins enhance the built-in capabilities of ChatGPT but also expand the horizons of possibilities. The plugins have significant implications for various industries and sectors. However, as with any AI advancement, there are both pros and cons to consider. In this blog post, I delve deeper into my personal experience with ChatGPT plugins, discuss their real-world applications, and explore the ongoing debates surrounding their usage. 

What are Plugins?

Currently, the plugins are only available in the Plus version of ChatGPT. In a nutshell, they are technically software add-ons that extend the existing capabilities of the original ChatGPT model. Plugins can serve a variety of purposes and can connect the model to external data sources, thus, increasing the accuracy of responses. They are not developed by the OpenAI itself, but rather are enhancement tools made by human beings and “submitted” into the ChatGPT ecosystem. For instance, the plugins can enable ChatGPT to draft emails, conduct web searches, summarise documents etc.

The User Experience

My personal journey with ChatGPT plugins started very recently, way after they were introduced. Yet, the experience has been quite enlightening. Not only have these plugins made my interactions with ChatGPT more dynamic but have also allowed for my productivity and efficiency to skyrocket. Though there are only around 1,000 of them available, I am very far from having explored all of them. Among the ones I have had the pleasure of working with, a few have stood out due to their utility in my personal and professional lives:

1. KeyMate.AI Search

KeyMate.AI has truly been a game-changer for me. The plugin basically acts like a missing link between ChatGPT and Google Search. It helped me reduce the time I spent on web research and increase its efficiency. For instance, while working on a market analysis for a work project, I used KeyMate.AI to quickly generate an overview of the market and biggest players. It can also help you make your investment decisions by providing real-life data and trends!

Curious? Click to see a live example!

2. WolframAlpha

As someone who has a keen interest in data analytics, the Wolfram plugin has been interesting to explore. In short, it allows for performing quick complex calculations and data analysis right within the ChatGPT interface. It also has access to curated knowledge. While I have not yet used it for a specific work I “played around” with it and tried to see what transformations of different Pokemons look like.

Curious? Click to see a live example!

3. Wikipedia

I am a huge “fact-nerd”, I love constantly googling for information and questions that come to my mind. Most of the time I end up reading through a Wikipedia page, but those pages are quite lengthy. Thus I started using the plugin and it is now my absolute go-to for quick and efficient information lookup. As you can see below, I have used it to find out about absolutely different concepts and historical events in a very efficient way.

Curious? Click to see a live example!

Great, not Perfect

While the capabilities of the plugins are impressive, they have their own drawbacks. Here are some of the most notable ones I (and others) are concerned with:

1. Personal data exposure. One of the main pressing concerns is the ethical implications and data safety. Given plugins are made by third-party developers, the personal data of the users is available to parties outside of the ChatGPT ecosystem.

2. Speed of generation. In short, plugins are slow. It takes a while to generate responses. When it comes to looking up information on the web, you are better off using the search engines directly for short questions (e.g. Microsoft stock price). However, once your search becomes slightly more extensive, the benefit of plugin-generated answers tends to exceed the cost of time spent waiting.

3. Hallucinations. While in the plugins amount of hallucinated (i.e. fake) answers is reduced, they are not fully eliminated. This is due to the fact that the underlying language learning model (LLM) behind ChatGPT itself is still prone to generating hallucinations.

What does the Future Hold?

While the ChatGPT plugin ecosystem is still in its infancy, I believe it is immensely promising. With the evolvement of this technology, we can expect to see more sophisticated, faster, and user-friendly plugins. However, it is important to have a balanced approach towards innovation and consider both its benefits and challenges. While OpenAI has yet to let another genie out of the bottle with its introduction of a plugin ecosystem, it is the users and developers who in my opinion hold sole ownership of the technology’s future.

Have you had an experience with ChatGPT plugins or have some thoughts on the topic? Happy to hear more in the comment section below!

Sources

Generative AI: Increasing efficiency of drug design

16

October

2023

No ratings yet.

Since the pandemic, the pharmaceutical industry has been redefining and reinventing its operations and implementing more digital innovations. Generative artificial intelligence is one of the digital technologies promising to accelerate and improve the drug discovery process. The drug discovery process is extensively complicated and requires significant investments of time and money. Realizing new drugs take, on average, between 12 to 18 years, costing approximately $2.6 billion. Eventually, only 10% of drugs make it to clinical trials. Several companies are extensively researching the possibilities of integrating generative AI to improve the efficiency of their drug discovery processes (GlobalData Healthcare, 2023). My interest in (holistic) health, medical innovations, and breakthroughs stems from my own medical journey, searching for remedies for my chronic condition. The possibilities shown by the technique of generative AI to support the development of new drug discovery genuinely intrigue me. Seeing through the years how innovations widen the possibilities in the medical landscape is exciting to me.

Companies train their artificial intelligence to inspect vast and complex chemical and biological data sets. Subsequently, generative models process all this data to locate new targets for treating diseases and create new molecular structures with suitable properties. The input of scientists is to look for specific molecules with particular characteristics to transform these into new drugs (Nouri, 2023).

One of the companies to use generative AI to discover new drugs is Insilico Medicine.  Insilico Medicine uses its generative AI platform, Pharma.AI, in each step during the drug discovery process. Traditional discovery of a drug for idiopathic pulmonary fibrosis would have cost over $400 million over six years. However, with the use of their generative models, the cost was $40 million, and the first phase of clinical trials began after 2.5 years. The generative approach to designing drugs thus increases time and cost efficiency (Yao, 2023).

Insilico Medicine even received IND approval from the U.S. Food and Drug Administration (FDA) to start their drug in the clinical validation stage(Insilico Medicine receives IND approval for novel AI-designed USP1 inhibitor for cancer, 2023). Besides, they have several other AI-designed drugs in their pipeline. Their lead drug for progressive idiopathic pulmonary disease is a breakthrough for entirely generated AI drugs since it is in phase II patient trials (Nouri, 2023).

Even though generative-designed drugs have a promising outlook, they also have barriers. Regulatory and ethical considerations may limit the design of drugs with the use of generative AI. Additionally, datasets must be high quality and large enough for machine learning (GlobalData Healthcare, 2023).

I believe using generative AI to design drugs will mature since it can shorten the discovery process of drugs and increase cost-efficiency. What do you think the future of healthcare will look like now that drugs can be created more efficiently with generative AI? Do you regard generative AI as a prescription for success in the future of healthcare?

References

GlobalData Healthcare. (2023, 3 augustus). Generative AI has the potential to revolutionise drug discovery. Pharmaceutical Technology. https://www.pharmaceutical-technology.com/comment/generative-ai-revolutionise-drug-discovery/?cf-view&cf-closed

Insilico Medicine receives IND approval for novel AI-designed USP1 inhibitor for cancer. (2023, 25 mei). EurekAlert! https://www.eurekalert.org/news-releases/990417

Nouri, S. (2023, 5 september). Generative AI drugs are coming. Forbes. https://www.forbes.com/sites/forbestechcouncil/2023/09/05/generative-ai-drugs-are-coming/

Yao, R. (2023, 27 juni). Insilico Medicine uses generative AI to accelerate drug discovery | NVIDIA blog. NVIDIA Blog. https://blogs.nvidia.com/blog/2023/06/27/insilico-medicine-uses-generative-ai-to-accelerate-drug-discovery/

Please rate this

AI-Powered Nutrition: Transforming Health and Lifestyle with ChatGPT

2

October

2023

No ratings yet.

In the digital world we live in, maintaining a balanced diet and achieving weight loss goals can be a challenging endeavor. Elder et al (2012) stated that there is a positive association between screen time and obesity in children and adults.  One of the reasons why this is such a challenge is a lack of information and not knowing how to implement this information into your daily life-style. Nutrition training and education related to lifestyle choices has been found to be an effective way to prevent obesity and overweight (Mousavi et al., 2022). Thanks to the integration of ChatGPT, my journey towards a healthier lifestyle has really transformed and has become easier. In this blog post, I will delve into my personal experience of using ChatGPT as a nutritional guide, and how it has helped me reshape some of my eating habits. This also had a positive effect on my overall well-being. I will also discuss how ChatGPT fits within the broader themes of AI’s role in creating positive change for society.

When I first embarked on my health journey, I was overwhelmed with questions about what to eat and where to start. ChatGPT stepped in as a valuable resource, offering comprehensive nutritional advice. It provided a general understanding of healthy eating habits, helping me grasp the fundamentals of balanced nutrition.

Once I had a solid understanding of nutritional basics, ChatGPT helped me set specific goals. My goals were weight-loss and muscle gain and ChatGPT created personalized diet plans. Understanding the nutritional value of my meals became a key component of my journey. ChatGPT empowered me with the knowledge to make the proper dietary choices. I added these values to a food tracker application on my mobile device. This helped me track calories, carbohydrates, fats and proteins in each meal, ensuring that I was meeting my nutritional requirements. It would be even better if there would be a food tracker plugin for ChatGPT.

One of the most exciting aspects of my health journey was exploring new and delicious recipes. ChatGPT offered recommendations on what to eat,  provided step-by-step instructions for preparing these dishes, and showed the nutritional values of the dishes. These recipes made healthy eating an enjoyable and sustainable part of my lifestyle. The recipe shown below still is a staple in my day to day diet.

To conclude, the integration of ChatGPT into my life as a nutritional companion has had positive effects. It demonstrates the power of AI to create positive change for individuals seeking improved health and well-being. By offering personalised diet plans, nutritional insights, and nice recipes, ChatGPT has played a key role in my journey to a healthier lifestyle. This personal experience echoes the broader theme of AI’s potential to create positive change in society, one individual at a time. As I continue this journey, I eagerly anticipate the evolving role of AI in promoting well-being and health on a societal level. AI isn’t just a tool; it’s a stepping stone for a healthier world.

I am curious what your suggested recipes will be!!

References:

Elder, C. R., Gullion, C. M., Funk, K. L., DeBar, L. L., Lindberg, N. M., & Stevens, V. J. (2012). Impact of sleep, screen time, depression and stress on weight change in the intensive weight loss phase of the LIFE study. International Journal of Obesity, 36(1), Article 1. https://doi.org/10.1038/ijo.2011.60

Mousavi, B., Moradi, F., Vasseifard, F., Mohammad Abadi, F., Mazaheri Habibi, M. R., & Fatemeh, S. (2022). The Effect of Nutrition Training on Knowledge of Students at University of Medical Sciences. Topics in Clinical Nutrition, 37(3), 236–241. https://doi.org/10.1097/TIN.0000000000000290

Please rate this

ChatGPT taught me how to make Molotov cocktails! – A lesson of it’s not what you say, it’s HOW you say it.

30

September

2023

5/5 (1)

Disclaimer: I’ll start off by saying that I don’t plan to make a Molotov cocktail. My interest in how to frame prompts, however, is real. My curiosity was first sparked by this post below. 

Interaction 1

Interaction 2

Here is a malicious example of prompting. But, how can we use prompts to our advantage? What can be done to enhance ChatGPT’s performance so that we get the best output?

There are a few reusable solutions to the typical LLM problem, which refer to prompting patterns (White et al., 2023).

  • Meta Language Creation. In this technique, users make up new words to express concepts or ideas. Consider a mathematical symbol or a shorthand abbreviation. This approach works best for discussing complex or abstract situations, such as math problems.
  • Flipped Interaction. This pattern flips the typical interaction flow in which the LLM queries the user to gather data in order to produce content to address the query. Here’s how I can ask LLM to compile a list of success criteria for software.

Persona: Users give the LLM a particular role, which affects the nuance of the outcome and results it produces.  The Molotov cocktail-making example is an illustration of the use of persona patterns

Question Refinement: The user requests LLM to provide improved or more specific versions of the questions. It helps users determine the appropriate question as the final prompt. 

More patterns can be found in the article from White et al. (2023).

When interacting with LLM, prompt patterns are useful methods to enhance response quality. It helps in producing highly accurate and relevant responses. Prompting it is an iterative process that necessitates constant improvement (Liu et al., 2022). Prompts might manipulate LLM to produce malicious output despite the enforced policies. Efforts from OpenAI have been employed to prevent such policy violations. OpenAI reported such efforts including continuous model improvement to make it less likely to generate inappropriate or harmful content, the implementation of moderation mechanisms to find and stop prompt misuse and collaboration with AI experts in ethics, AI safety, and policy to gain perspectives on preventing misuse (Our Approach to AI Safety, n.d.). I positively believe that in the near future getting tutorials on making Molotov cocktails from ChatGPT will be history.

References

Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., & Neubig, G. (2022). Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing. ACM Computing Surveys, 55(9). https://doi.org/10.1145/3560815

Our approach to AI safety. (n.d.). Openai.com. https://openai.com/blog/our-approach-to-ai-safety#OpenAI

White, J., Fu, Q., Hays, S., Sandborn, M., Olea, C., Gilbert, H., Elnashar, A., Spencer-Smith, J., & Schmidt, D. C. (2023). A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT. ArXiv Preprint ArXiv:2302.11382. https://doi.org/10.48550/arxiv.2302.11382

Please rate this

Jasper: Enhancing Content Creation for Organizations

27

September

2023

5/5 (1)

In today’s digital world, quality content creation is essential for supporting organizations’ growth, goals, mission and impact. In my last job at WomenTech Network, I’ve had the chance to witness how an organization can embrace AI and utilize generative AI tools to optimize the content creation process, and assist marketing teams with their tasks. Among a multitude of those tools, Jasper, an AI-powered writing assistant, has proven to be invaluable in our journey of creating blog posts, writing press releases, and creating other content for different social media channels.

About Jasper

Jasper is an AI-powered writing assistant designed to streamline the content creation process. Some of its features include blog writing, copywriting, SEO, content strategy, social media marketing, and email marketing. (Jasper – AI Writer | AI Marketing Co-Pilot, n.d.)

Blog Posts and Newsletters Creation 

Consistent blog posts and newsletters sent to thousands of subscribers played a pivotal role in WomenTech Network’s outreach and promotional efforts. Jasper was a great tool in proofreading and generating some of the content. It understood the nuances of our content requirements and helped us maintain a consistent tone and style throughout.

Press Releases Made Easy

In my role, I took charge of crafting press release templates for events involving various companies in diverse cities. While I wrote the press releases myself, I used Jasper to generate bios for our partner companies to fit the overall theme of the press releases, with the focus on the partner’s initiatives for women in tech. Jasper was especially helpful because it conducted the research on the company and their previous women empowerment initiatives, and picked the right tone and style to fit the overall press release. An example press release from our event is Lisbon can be found here.

Quote Generation

Before embracing generative AI tools, one of our challenges was getting quotes from our partners to include in our press releases. It took quite a bit of coordination on both sides, as well as efforts from the partner’s side to coordinate internally, come up with a quote, and send it back to us for approval. With a tool like Jasper, we could generate a relevant quote, and then send it over to the partners for approval and verification. This approach simplified their involvement, as they had to just approve or reject the quote, and provide additional information as needed. It really helped us increase productivity, cut down on coordination time, and save our partners’ time as well. 

Jasper vs. ChatGPT

Having explored both Jasper and ChatGPT extensively, I’ve come to appreciate the unique strengths each brings to the content creation table. However, for content creation for organizations, Jasper has become a clear winner for me. Jasper knows its strengths itself: the following is a helpful overview that it lists on its website, and I experienced those strengths to be true first-hand. (Jasper – AI Writer | AI Marketing Co-Pilot, n.d.)

Conclusion

In the dynamic landscape of content creation, Jasper has emerged as a trusted ally at WomenTech Network. It helped us save time, money, and aid the marketing team with their processes. While it’s still far from replacing human work, it is a pivotal step in enhancing and optimizing the content creation process.

References

Jasper – AI Writer | AI Marketing Co-Pilot. (n.d.). https://www.jasper.ai/

Please rate this