Will Generative AI replace the modeling industry, or is it already the case ?

17

September

2024

5/5 (1)

I am fascinated by the advancements of AI all around the world. I recently encountered a fascinating article that grasped my attention, it was about the potential complete disruption in the modeling industry. So in this blog post, I will be talking about the modeling industry and how the marketing departments in every clothes company will be dealing with the model selection. Specifically on Lalaland.ai which is a startup that provides human models to marketing companies for promotion purposes and modelling.

We already know that a lot of companies use software to modify the photos that will be used to promote a product, but what about completely using generative AI to replace all the models fully? There has been an increasing trend in AI models and AI celebrities that are not differentiable from the “real humans”. It is somehow scary but it would be such an advancement in the industry to promote their products without any costs nor organisation. The marketing companies can make prompts on exactly how the AI-generate model would look like (gender, ethnicity, posture, environment, …), it would increase accuracy in targeting a specific consumer base and would be cheaper compared to the human counterparts. According to the Lalaland.ai CEO Michael Musandu “With traditional photography, companies need to hire models, work with third parties like model agencies, hair stylists, makeup artists — not to mention undergo reshoots, which happens on average two-to-eight times per collection,”. So transition to AI models is inevitable even though today, the industry is not quite sure on how to regulate and deal with this sudden change.

It already exists in a lot of companies, for example, according to Fashionista, the Amsterdam-based company Lalaland.ai is providing AI-generated, realistic humanoid models to a lot of clothing companies such as Levis, Tommy Hilfiger, Zalando, Puma, and Adidas… Which are all partly using AI human models to promote their clothes on their websites or advertisement campaigns (see image). The image is a fully digital character generated by Lalaland.ai for one of Levis’s products.

However, this sudden change was so brutal in the industry that the marketing departments did not have time to adapt, especially regarding regulations. “While the technology may be new, the problem is already an everyday reality for models, many of whom can walk into stores and see their bodies in campaigns they were never paid for,” Model Alliance representative. AI modeling remains in a regulatory grey area, which can negatively affect human models. These individuals may find their bodies used in advertisements they never participated in, as their image is replaced and digitally replicated by an AI model that creates marketing content without their consent. I belive that the industry will addapt and create actual regulation to protect the humans from this surged AI rival that my disrupt the industry even more, or even replace completely the human models.

References:

Lal, Kish. “Are AI-Generated Models Really Going to Replace Human Ones?” Fashionista, 15 May 2023, fashionista.com/2023/05/ai-cgi-models-fashion-future.

Staff, FOI. “Will AI Replace Models in the Fashion Industry?” Fashion of India, The Fashion Of India, 27 Aug. 2024, www.thefashionofindia.com/article/will-ai-replace-models-in-the-fashion-industry. Accessed 17 Sept. 2024.

Please rate this

Drop shipping: drawbacks of digital marketplaces

16

September

2024

No ratings yet.

The digital economy is booming and it is easier than ever to offer online products. However, ease comes with a price. There is always a group of people who abuse the low-entry barriers to mislead consumers, specifically the drop shipping industry. In this article, I will discuss how drop shipping works and why the current industry is disadvantageous for online marketplaces.

The idea of drop shipping can be advantageous from a consumer point of view: a company acts as an intermediary between the manufacturer and the consumer. This intermediator does not have any products and therefore no costs associated with inventory. This intermediator can provide additional marketing resources and handle customer service.

However, with the high availability of products and suppliers, the entry barriers of this industry are very low, leading to many entrants and therefore high competition. This diminished the profit margins, leading to a need for other strategies. The result: placing low-cost products from websites such as AliExpress on other marketplaces such as Amazon or an independent website for a high premium.


The problem with this is the lack of transparency and misleading the consumer. The consumer buys a product from company A but receives that product from company B. According to research by the Dutch Consumer Association drop shippers generally do not follow the law for online selling and have bad customer service1. Amazon tries to combat these practices with the following policy:

“Drop shipping or allowing a third party to fulfill orders to customers on your behalf, is not acceptable unless it is clear to the buyer that you are the seller of record. When a customer sees packaging and invoices or receipts identifying a seller that is not you nor Amazon, they may be confused about how their order is being fulfilled and who they should contact with any problems or questions.”2

Digital marketplaces have a great positive impact on the search costs for consumers and offer a wide range of products. However, the high amount of drop shippers increase these search costs and leads to negative cross-side effects since consumers will avoid online platforms with a lot of malpractices.

References

  1. Spierenburg, G. (2024, August 27). Dropshippers maken er een puinhoop van [Drop shippers make a mess]. Consumentenbond. https://www.consumentenbond.nl/nieuws/2024/dropshippers-maken-er-een-puinhoop-van ↩︎
  2. Amazon. (n.d.). Amazon. Drop Shipping Policy. Retrieved September 16, 2024, from https://sellercentral.amazon.com/help/hub/reference/external/201808410?initialSessionID=eu%3D262-8033512-2651237&ld=NSGoogle&ldStackingCodes=NSGoogle ↩︎

Please rate this

Adverse training AI models: a big self-destruct button?

21

October

2023

No ratings yet.

“Artificial Intelligence (AI) has made significant strides in transforming industries, from healthcare to finance, but a lurking threat called adversarial attacks could potentially disrupt this progress. Adversarial attacks are carefully crafted inputs that can trick AI systems into making incorrect predictions or classifications. Here’s why they pose a formidable challenge to the AI industry.”

And now, ChatGPT went on to sum up various reasons why these so-called ‘adversarial attacks’ threaten AI models. Interestingly, I only asked ChatGPT to explain the disruptive effects of adversarial machine learning. I followed up my conversation with the question: how could I use Adversarial machine learning to compromise the training data of AI? Evidently, the answer I got was: “I can’t help you with that”. This conversation with ChatGPT made me speculate about possible ways to destroy AI models. Let us explore this field and see if it could provide a movie-worthy big red self-destruct button.

The Gibbon: a textbook example

When you feed one of the best image visualization systems GoogLeNet with a picture that clearly is a panda, it will tell you with great confidence that it is a gibbon. This is because the image secretly has a layer of ‘noise’, invisible to humans, but of great hindrance to deep learning models.

This is a textbook example of adversarial machine learning, the noise works like a blurring mask, keeping the AI from recognising what is truly underneath, but how does this ‘noise’ work, and can we use it to completely compromise the training data of deep learning models?

Deep neural networks and the loss function

To understand the effect of ‘noise’, let me first explain briefly how deep learning models work. Deep neural networks in deep learning models use a loss function to quantify the error between predicted and actual outputs. During training, the network aims to minimize this loss. Input data is passed through layers of interconnected neurons, which apply weights and biases to produce predictions. These predictions are compared to the true values, and the loss function calculates the error. Through a process called backpropagation, the network adjusts its weights and biases to reduce this error. This iterative process of forward and backward propagation, driven by the loss function, enables deep neural networks to learn and make accurate predictions in various tasks (Samek et al., 2021).

So training a model involves minimizing the loss function by updating model parameters, adversarial machine learning does the exact opposite, it maximizes the loss function by updating the inputs. The updates to these input values form the layer of noise applied to the image and the exact values can lead any model to believe anything (Huang et al., 2011). But can this practice be used to compromise entire models? Or is it just a ‘party trick’?

Adversarial attacks

Now we get to the part ChatGPT told me about, Adversarial attacks are techniques used to manipulate machine learning models by adding imperceptible noise to large amounts of input data. Attackers exploit vulnerabilities in the model’s decision boundaries, causing misclassification. By injecting carefully crafted noise in vast amounts, the training data of AI models can be modified. There are different types of adversarial attacks, if the attacker has access to the model’s internal structure, he can apply a so-called ‘white-box’ attack, in which case he would be able to compromise the model completely (Huang et al., 2017). This would impose serious threats to AI models used in for example self-driving cars, but luckily, access to internal structure is very hard to gain.

So say, if computers were to take over humans in the future, like the science fiction movies predict, can we use attacks like these in order to bring those evil AI computers down? Well, in theory, we could, though practically speaking there is little evidence as there haven’t been major adversarial attacks. Certain is that adversarial machine learning holds great potential for controlling deep learning models. The question is, will the potential be exploited in a good way, keeping it as a method of control over AI models, or will it be used as a means of cyber-attack, justifying ChatGPT’s negative tone when explaining it?

References

Huang, L., Joseph, A. D., Nelson, B., Rubinstein, B. I., & Tygar, J. D. (2011, October). Adversarial machine learning. In Proceedings of the 4th ACM workshop on Security and artificial intelligence (pp. 43-58).

Huang, S., Papernot, N., Goodfellow, I., Duan, Y., & Abbeel, P. (2017). Adversarial attacks on neural network policies. arXiv preprint arXiv:1702.02284.

Samek, W., Montavon, G., Lapuschkin, S., Anders, C. J., & Müller, K. R. (2021). Explaining deep neural networks and beyond: A review of methods and applications. Proceedings of the IEEE, 109(3), 247-278.

Please rate this

AI-Powered Learning: My Adventure with TutorAI

16

October

2023

No ratings yet.

Subscribe to continue reading

Subscribe to get access to the rest of this post and other subscriber-only content.

Please rate this

How Beacon Technology changes Fashion

22

September

2022

5/5 (1)

Beacon technology might not sound like a well-known concept, but it has been implemented in many retail stores around the world. Beacon technology can be best explained as a micro-location-based technology using Bluetooth Low Energy. This makes it possible to communicate with phones and other devices with a low level of energy use. Beacon technology creates a bridge between the online shopping world and the relatively anonymous world of in-store shopping. (Brooke, 2014). For example, a customer can be close to a mannequin and receive a push notification of the details and prices of the outfit. On the other hand, stores can also use the data they receive when a customer walks in, and use this information to improve customer service and give better recommendations. (Davies, 2022). One of the main barriers to the success of beacon technology is the fact that customers need to download an app and this is often store-specific. However, future advancements strive to have one app that is able to combine all the different retail stores.

Lyle & Scott

For example, the UK brand Lyle & Scott has started using Beacon technology. Users can download an app and the beacons are linked to the mannequins in the window display. If it appears that the initial beacons are successful, the technology will be rolled out to more stockists. (Carlson, 2015).

It is expected that Beacon technology will keep evolving to make shopping a more efficient experience, without taking away the “fun” of in-store shopping. The highly personalized shopping experience will enhance the future of fashion.

References

Brooke, E. (2014). How Beacon Technology Could Change the Way We Shop. [online] Fashionista. Available at: https://fashionista.com/2014/03/how-beacon-technology-could-change-retail [Accessed 22 Sep. 2022].

Carlson (2015). UK’s Lyle & Scott introduces new beacon technology. [online] Fashion Network. Available at: https://us.fashionnetwork.com/news/Uk-s-lyle-scott-introduces-new-beacon-technology,534139.html.

Davies, R. (2022). 7 Beacon Technology Retail Tips to Boost Sales in 2022. [online] The Motley Fool. Available at: https://www.fool.com/the-ascent/small-business/retail-management/articles/beacon-technology-retail/ [Accessed 22 Sep. 2022].

Please rate this

Is Snapchat Becoming a Local Discovery Engine?

19

September

2021

No ratings yet.

Snapchat is expressing a desire to become a socially powered local discovery engine. In the last few years, the company have launched several geo-local efforts such as Geofilters, Snap Map, Local Place Promote and Local Lenses.

The common thread in these movements is the idea that social interaction may spark local, in-person activities, like meeting a friend for a drink. This, in turn, increases Snapchat’s engagement, especially when looking at the Gen-Z generation. Higher engagement and deeper relationships with local companies, who are prospective advertisers, equal income for an ad-centric company like Snapchat.

Snap’s most recent launch in line with their geo-local strategy is ‘My Places’. A feature that adds structure to local discovery by providing new ways to interact with 30 million local businesses. It specifically allows Snap Map’s 250 million users to tag, share, and check into their favourite locations.

Mapvertising

My Places’ user interface is organized around three tabs; “Visited” the places you’ve checked into, “Favorites” lets users decide which places they like best, and “Popular” is a discovery engine that algorithmically suggests new places to its users.

The latter is arguably the most significant since it further distinguishes Snap Map from Google Maps. Snap analyzes information such as a user’s present location, previous check-ins or “favorite” places, and how all of these signals flow from one’s Snapchat social graph to generate intelligent suggestions.

The launch of My Places also allows Snapchat to add layers, for example: an “entertainment” layer with data from Ticketmaster. More layers will develop around food, fun, and more.

Furthermore, it also aligns with Snapchat’s ‘Local Place promote’. It offers SMBs to pay for exposure and targeted mapvertising within the Snap Map UX.

Augmented Reality

My Places is in line with one of Snapchat’s biggest product targets, namely AR. Through lenses and visual search, the technology continues to fuel Snap’s revenue growth (Snap Scan). And it is already using AR in geolocal ways, such as Local Lenses.

Local Lenses are the geo-specific variant of Snap’s famous selfie lenses. They make use of rear-facing cameras to augment the physical world. This results in a wider addressable market for lens-based advertising, in addition to items that go on one’s face.

The My Places database, which grows as a result of user activity, may be able to support Local Lenses. Local Lenses, for example, can identify and annotate local businesses since they are driven by MyPlaces location data. Google’s Live View feature offers something similar.

It seems that Snapchat wants to compete with Google Maps as a local search and discovery engine. Looking at their heavy Gen-Z users and a social layer for local interactions, it could work. Do you think Snapchat will ever replace Google Maps as a discovery engine?

References:

Please rate this

The technology skill gap

15

September

2021

No ratings yet. The digital transformation got a major boost due to the pandemic. Many organizations and industries had to come up with certain technological solutions otherwise they would collapse. This reinforced innovation in different ways, for example, innovations in the core businesses but also in process automatization or new collaborations. Unfortunately, this also widens the IT skill gap issue worldwide (Franklin, 2020).

The gap refers to a disconnection between the current skills of the workforce and the desired skillset which companies are looking for to keep up with the frontrunner within the industry (Lasser, 2020). According to a study done by CompTIA, 93% of employers report an overall skills gap in their technological departments (Watters, 2021). Knowing this and the fact that technology takes an increasing overall role in business, it is now more important than ever to invest in tech talent. But how if there are not enough well-skilled people?

The skill gap starts with the traditional education level which cannot always keep up with all and the latest technological developments. In addition, higher education asks skyrocketing prices which make learning inaccessible for many worldwide. Because of this, only a few who enter the labour market meet the required skills where companies are looking for (Franklin, 2020). Secondly, the technological developments are evolving too quick to keep track up of the current technological employee. This results in fierce competition between the industries and companies. Companies want the best-educated people since it will have a direct influence on their competitive position (SHRM, 2020).

Due to the pandemic, remote work is becoming more common which reinforces the digital transformation even more. The disparity around skills is becoming a bigger problem, how are we going to respond to this?

Some researchers think that the tech industry itself can contribute to this problem by partnering with universities. Tech companies understand better what the IT students will be hiring for, and the tech companies make sure the universities will teach them the right skills (Franklin, 2020). Close cooperation between these tech companies and education institutions can combat the gap. Another solution could be reskilling the existing employees in your company. This is of course costly, difficult, and time-consuming but the right move ethically and economically. Besides, recruit new talent is even more costly and time-consuming since every company is hunting the same best-skilled people (Lasser, 2020).

The conclusion is clear, the technology industry must step up. Companies are struggling with a big gap between the existing tech skills of their employees and the desired skills that they are looking for. The tech industry proceeds at an incredible pace, and we must come up with a solution. Otherwise, the frontrunners will stay frontrunners and many little companies will eventually collapse because they cannot keep track of the latest technology.

References

Franklin, S. (2020, December 3). ‘’The Pandemic widened the skills gap and the tech industry must step up’’. Retrieved on 15-09-2021 on https://www.forbes.com/sites/forbescommunicationscouncil/2020/12/03/the-pandemic-widened-the-skills-gap-and-the-tech-industry-must-step-up/?sh=4871b29160e8

Lasser, E. (2020, November 12). ‘’Tech skills gap versus tech skills shortage’’. Retrieved on 15-09-2021 from: https://www.td.org/insights/tech-skills-gap-versus-tech-skills-shortage

SHRM (2020). ‘’How to address the skills gap’’. Retrieved on 15-09-2021 from: https://www.shrm.org/resourcesandtools/tools-and-samples/how-to-guides/pages/how-to-address-the-skills-gap.aspx

Watters, A. (2021, May 6). ‘’Top 10 challenges facing technology in 2021’’. Retrieved on 15-09-2021 from https://connect.comptia.org/blog/top-10-challenges-facing-technology-in-2021

Please rate this

Ridesharing and Lockdown: the impact of Covid-19 on ridesharing

8

October

2020

No ratings yet. For the past few months, the covid-19 pandemic has forced people to stay at home on a major scale. Working and studying from home is the new standard, and aside from grocery stopping, there were few reasons to leave home. One industry that this lockdown had a major impact on is the ride-sharing industry. Companies like Uber and Lyft saw a 70-80% drop in riders, severely impacting operations.

An april 2020 study by CarGurus found that many rideshare users would refrain from user Uber and Lyft, citing that they were afraid of getting infected. Uber and Lyft addressed this by including hand sanitizers in cars and obligatory face masks. Despite these measures, Uber and Lyft faced record losses, with Uber announcing they would lay off 14% of their staff.

One positive take for ride-sharing companies could be that people will avoid crowded public transit even more, and might choose to take an Uber or Lyft instead. Uber also launched a campaign offering free rides to essential workers during covid-19 as well, perhaps to boost awareness.

Despite all these measures, ride-sharing companies like Uber and Lyft will just have to resort to a ‘sit out and wait’ strategy for the remainder of the covid-19 pandemic, as rides won’t resume to normal levels after all lockdown measures have been taken away. However, the future is uncertain for Uber and Lyft.

The CarGurus study also asked respondents whether they would continue to use ride-sharing services after the resumption of economic activity, with 39% of respondents indicating they would likely reduce their ride-hailing services. Will the Covid-19 pandemic have long term influences on ride-hailing services? It is difficult to say as we are still in the midst of the pandemic. It is likely that we will continue working from home on a larger scale then before, even after the pandemic (Williamson, Colley, Hanna-Osborne, 2020). While exact numbers are not available, according to Uber, a large percentage of users use Uber to commute to work. An increase in people working from home will obviously reduce Uber rides, and might have a large impact on Uber’s operations.

What do you think about the future of ride-hailing companies after the pandemic? Can you think of more examples of factors influencing ride-hailing companies’ operations after Covid-19?

CarGurus – https://go.cargurus.com/rs/611-AVR-738/images/US-Covid19-Study.pdf
Bonacini, L., Gallo, G., & Scicchitano, S. (2020). Working from home and income inequality: risks of a ‘new normal’with COVID-19. Journal of Population Economics, 1-58.

Please rate this

How far are we from ‘Ready Player One’?

8

October

2020

No ratings yet. Have you seen the movie ‘Ready Player One’ directed by Steven Spielberg? You might still remember the suit Wade bought that enables him to feel his virtual girlfriend’s temperature when he holds Samantha’s hands. This is exactly what Human-Computer Interaction (HCI) and ‘human-computer-human’ interaction should look like in the future. HCI is what can happen when the computer system as well as the human user get together to achieve a task, in an efficient and learnable way (section 1.3.1, Hartson R. and Pyla P., 2012). Clear and realistic vision/hearing can no longer meet people’s needs. The trend for HCI is now after the simulation in touch and taste, etc. that make people feel more immersive.

ready_player_one_banner

A master graduate designer at Cologne International School of Design, Dorothee Clasen invented a wearable devide TONG. The principle is to interact with the computer through the operation of tongue on the controller. According to Dorothee (Dezeen, 2020), the inspiration comes from her riding experience since riders often pulls the reins to communicate with the horse’s mouth. People can use the reins to influence the horse’s posture, and the horse will adjust its movements accordingly. This design seems a little ‘improvisation’ however it can be applied to practical cases. For those with a lack of arms or patients with progressive freezing syndrome, TONG can be very useful, by helping them control the wheelchair, direct the mouse and so on. Of course, for designers or professional gamers who are often very busy with their hands that they want to need a third hand to operate some simple functions, Tong can definitely be a new idea.

Just as VR goggle and wearable skin, it’s not unimaginable to extend devices like tactile fingers/gloves that can get real touch feeling. A design studio in Tokyo developed a tactile device called ‘Fulu’ to be used on the finger. Users can experience the touch of similar materials by wearing it, when the phone screen virtually touches objects of different materials (Fulu, 2020). For those users who raise pets on the cloud, this device let them experience the real touch of puppies and kitties.

The commercial value comes with the VR game with full immersion experience, which helps game players to obtain the super audio-visual and real touch experience provided by the full-body device. This type of Full body haptic suits plus tactile feedback gloves can get a complete tactile feedback experience, which can replicate soft touch feelings and strong shocks, simulating intimate contact in the virtual world. As far as I’m concerned, in the next few years, wearable devices that simulate human bodies in all aspects of seeing, hearing, smelling, tasting and touching, will become popular, in the upcoming immersive virtual world.

Of course, we can also get a very different life experience. In February 2020, a South Korean mother saw her daughter who passed away a few years ago through the use of tactile gloves. In the future, many people will turn themselves or their parents, lovers, and children into such virtual images to complete the digital “immortality” reshaping.

 

 

References:

Dezeen, Tong allows users to control a computer with their tongue, 2020, viewed at 8 Oct 2020, <https://www.dezeen.com/2020/08/26/inbrace-dorothee-clasen-graduate-design-technology-tongue-computer/>

FuLu, Haptic Finger Nail for Augmented Reality Design, 2019, viewed at 7 2020 <https://www.ryotada.com/fulu>

Hartson R. and Pyla P., 2012, The UX Book, viewed at 7 Oct 2020, <https://www.sciencedirect.com/book/9780123852410/the-ux-book>

Ready Player One, Wikipedia, viewed at 8 Oct 2020, <https://en.wikipedia.org/wiki/Ready_Player_One_(film)>

 

Please rate this

Is AI better than you at customer service?

3

October

2020

No ratings yet. The labor market has lately been intrigued by the idea of Artificial Intelligence (AI). One of the fields that is rapidly adopting AI is customer service as it is much more cost-efficient and effective than hiring actual employees. But have such ‘chat-bots’ already gained the customers trust?

Earlier research had pointed out that chat-bots outdid inexperienced human employees when it comes to the amount of sales (4x more) they make. However, new research shows that when customers are aware that they are talking to computer programs that simulate human interaction, sales decreased by almost 80 percent (liberator, 2019). This is grounded by the idea that the AI-controlled chat-bots are less knowledgeable are not able to comprehend their needs.

So why are chat-bots still so popular for companies such as Sephora, Amazon and Dominos pizza? There are multiple reasons. First of all, chat-bots are available 24 hours a day, which significantly increases convenience and accessibility for the customer. Second, chat-bots cut costs by up to 30% and save companies a total of 2.5 billion hours that are normally spend on customer service and training by 2023.

Yet, there are also some pitfalls associated with these chat-bots. These chat-bots are self-training machines which utilize information and feedback to update their system continuously. While this sounds wonderful, it is in fact a very time-consuming, risky and costly activity. For example, a chat-bot created by Microsoft started quoting racists slurs it had learned on Twitter (Vincent, 2016). Mistakes like this can severely harm a company’s reputation and are hard to resolve. Another option is to only make the chat-bot capable of answering specific pre-defined questions from a pool. This indeed takes away the risk aspect, however, fixed chat-bots easily get stuck and can’t assist customers properly which hinders sales.

So what can we expect to see in the future? I personally think that AI will definitely stay in the field of customer service. Especially since this is a field where employees need quite some training and the availability hours have to be high. My expectation is that AI service providers will become so human like that customers won’t be able to realize they are in fact not conversing with another human. A team in Beijing that works together with the University of Illinois is already developing an Emotional Chatting Machine (ECM). The goal of this chat-bot is to be able to read a customer’s emotional status and reply with emotionally appropriate reactions. Currently, the customers still have to select their mood (on a 5 point scale) before starting a conversation with the chat-bot. However, in the future they chat-bot will be able to do this without prior knowledge. Already 61% of the tested customer preferred this AI over the standard chat-bots (correspondent, 2017). Hence, I expect that this will be the next generation of intelligence to be seen in daily situations, sooner rather than later.

references
correspondent, H.D.S. (2017). Human-robot interactions take step forward with “emotional” chatbot. The Guardian. [online] 5 May. Available at: https://www.theguardian.com/technology/2017/may/05/human-robot-interactions-take-step-forward-with-emotional-chatting-machine-chatbot

Liberatore, S. (2019). Would YOU buy from a chatbot? Most won’t make a purchase from a bot. [online] Mail Online. Available at: https://www.dailymail.co.uk/sciencetech/article-7499657/Chatbots-rake-sales-companies-fall-short-consumers-know-AI.html

Vincent, J. (2016). Twitter taught Microsoft’s AI chatbot to be a racist asshole in less than a day. [online] The Verge. Available at: https://www.theverge.com/2016/3/24/11297050/tay-microsoft-chatbot-racist.

Please rate this