Is the drive among enterprises across all regions and industry sectors to adopt generative AI solutions at scale living up to the hype or is the process moving at a slower and steadier pace?

The latest McKinsey Global Survey on AI describes 2023 as generative AI’s “breakout year”.

One-third of the survey respondents say their organisation already uses generative AI for at least one function, 40% of AI adopters are expecting to invest in more AI solutions because of the potential of generative AI and 28% say generative AI use is already on the board’s agenda.

A few of the speakers at the upcoming Generative AI Week (Oct 17-18, Atlanta) share their insights on the scale up process.

Get your copy to learn more in 6 simple steps >>>

Generative AI has captured the imagination of many, promising to revolutionize various industries through its ability to create new content, data, and even generate realistic images. However, despite the hype surrounding its disruptive potential, most enterprise businesses are cautiously treading the waters, considering the technology's risks to outweigh its immediate benefits.

For startups and challenger brands, the allure of Generative AI lies in its potential to provide them with a competitive edge. They have been quick to adopt the technology and explore its applications in different areas of their operations. On the other hand, incumbent businesses, with more at stake, are approaching Generative AI with a more measured and conservative stance. Their concerns are valid, as high-profile cases of hallucinations, intellectual property (IP) leakage, and cyber-security breaches have raised red flags.

Notwithstanding these reservations, there are exceptions to this conservative approach. Zurich Insurance, for instance, is an example of an incumbent business that is actively embracing Generative AI across multiple business lines. Equally, Booking.com have recently integrated Generative AI into their trip planner to enhance personalisation. Both businesses have recognized the potential benefits of the technology in driving efficiency, identifying opportunities, and enhancing the customer experience.

Use Case Exploration

Expanding the realm of Generative AI unveils a myriad of novel use cases that extend beyond its current applications. One such domain is drug discovery, where Generative AI is transforming the process of identifying potential drug compounds. By analyzing vast datasets of chemical structures and biological interactions, this technology assists researchers in generating novel molecules with the potential to combat various diseases more efficiently than traditional methods.

Copywriting is another burgeoning use case for Generative AI. Beyond marketing, this technology is now being harnessed to craft persuasive and engaging content across diverse industries. From generating compelling product descriptions to drafting captivating news articles, Generative AI tools are proving their adeptness in producing coherent and contextually relevant written material.

In the field of intellectual property, Generative AI is revolutionizing the patent application process. It can analyze existing patents, scientific literature, and technical documentation to generate comprehensive and precise patent applications. This not only accelerates the patent filing process but also enhances the quality of submissions by identifying potential gaps and opportunities.

Simulating real-world scenarios for training purposes is another innovative application. Generative AI can create lifelike simulations for training professionals in high-risk fields, such as aviation and surgery. These simulations provide a safe environment to practice complex tasks, improving skill acquisition and boosting performance in critical situations.

Furthermore, Generative AI is being harnessed for eco-friendly architecture and design. By inputting parameters like sustainability goals and material constraints, it can generate architectural plans that prioritize environmental impact alongside aesthetics and functionality.

Summarization and idea generation continue to be essential use cases. Summarization algorithms sift through voluminous text, distilling core ideas into concise summaries, aiding in efficient information consumption. Idea generation, on the other hand, catalyzes creativity by producing a multitude of innovative concepts based on minimal input.

Industries Leading the Way

In the landscape of Generative AI adoption, financial services and healthcare stand out as industries making rapid progress, in spite of the challenges they face as highlight regulated industries. In the financial sector, Generative AI is reshaping operations, from fraud detection to investment strategies. Its ability to generate synthetic data aids in stress testing financial models and optimizing risk management. In healthcare, the technology accelerates drug discovery through molecular generation and assists in medical image analysis, revolutionizing diagnosis and treatment planning.

Beyond these front-runners, several industries are quickly embracing Generative AI. Retail is leveraging it to customize marketing campaigns and enhance customer experiences, while manufacturing employs it for product design and process optimization. Additionally, the entertainment sector is employing Generative AI for content creation and special effects, redefining storytelling and visual experiences.

IP and Data Risks

As the landscape of Generative AI evolves at a rapid pace, the task of selecting the appropriate model takes on profound significance, shaping not only a company's technological landscape but also impacting its legal, ethical, and competitive positioning.

Within this multifaceted decision-making process, two critical concerns come to the forefront: the security of an organization's proprietary data and the risk of copyright infringement. The potential for AI-generated content to resemble existing copyrighted material raises valid legal concerns, necessitating a delicate balance between innovation and respecting intellectual property rights to avoid potential legal entanglements.

Moreover, the proliferation of AI-generated content intensifies the challenge of safeguarding sensitive data. As the volume of such content grows, the responsibility to ensure robust data privacy and protection becomes more intricate, requiring heightened diligence and comprehensive security measures.

Amid these intricacies, the choice of the right Generative AI model transcends mere technical considerations, evolving into a strategic decision that encompasses legal and ethical dimensions. By thoughtfully navigating these dual risks and opportunities, businesses can harness the boundless potential of Generative AI while minimizing risks, ensuring compliance, and solidifying their role at the vanguard of responsible and ethical innovation.

Cyber Security Challenges and Opportunities

The integration of generative AI in cyber security presents a double-edged sword. On one hand, it promises transformative potential; on the other, it poses critical challenges. Cyber security concerns are heightened due to the vast data requirements of AI model generation. Any security breach could lead to unauthorized access or sensitive data manipulation. To mitigate these risks, businesses must invest heavily in robust security measures to safeguard their AI infrastructure and the data it handles.

However, challenges bring opportunities. Generative AI can revolutionize threat detection by swiftly identifying patterns in security logs and network data. Automated incident response powered by AI can minimize human error and response times, effectively countering emerging threats. Vulnerability assessment through AI-driven simulations can proactively identify and address system weaknesses. The technology also offers enhanced user authentication and can play a crucial role in identifying and mitigating phishing attacks.

Despite the challenges, the integration of generative AI in cyber security holds the promise of fortified defense strategies and a more secure digital environment. Organizations that navigate these challenges stand to lead in safeguarding the digital realm, shaping a future where AI and cyber security coalesce for the better.

Risk Management Frameworks

To address these concerns and navigate the complexities of Generative AI, businesses are turning to risk management frameworks. These frameworks enable them to assess the potential risks involved and develop strategies to mitigate them effectively. Choosing the right generative AI model is a pivotal initial step within this process.

Factors such as the specific task requirements, the size and quality of the available dataset, and the desired level of output specificity play a decisive role in model selection. For instance, text generation might necessitate a different model choice than image synthesis due to the distinct nature of data types. Balancing computational efficiency with the level of creative fidelity required is also crucial; state-of-the-art models often demand significant computational resources.

As the field rapidly evolves, staying well-informed about the latest advancements and best practices is essential for making informed decisions that drive successful outcomes. Collaboration between domain experts, AI researchers, and ethicists further enriches the decision-making process, ensuring that not only technical aspects but also ethical considerations are thoroughly addressed. In this dynamic landscape, embracing flexibility and adaptability in model selection approaches will be instrumental in harnessing the true potential of Generative AI.

Deciding on the Right Generative AI Model

In the rapidly evolving landscape of Generative AI, selecting the most suitable model has become a pivotal decision-making process for businesses. The speed at which advancements are occurring underscores the significance of this choice. Organizations find themselves at a crossroads, pondering whether to opt for a 'black box' solution such as open AI, harness the capabilities of open-source models for in-house applications, or delve into the realm of bespoke models trained on proprietary data. Each option presents its unique set of advantages and challenges.

The allure of a 'black box' solution like open AI lies in its ready-made functionality, enabling businesses to swiftly integrate cutting-edge technology into their operations. However, this convenience often comes at the cost of limited customization and potential dependence on external providers. On the other hand, open-source models empower organizations with the freedom to tailor the technology to their specific needs, fostering internal innovation. Nevertheless, this path necessitates significant technical expertise and resources.

Intriguingly, the option of developing a bespoke model training on proprietary data offers the prospect of tailor-made solutions finely tuned to an organization's unique requirements. This avenue, though promising, demands substantial investments in data collection, model training, and ongoing maintenance.

Ultimately, the choice hinges on the intricate balance between immediacy, customization, and resource allocation. Each path holds potential, and businesses must carefully evaluate their objectives, resources, and long-term strategic vision. In this ever-evolving landscape, the decision regarding the right Generative AI model is a pivotal one, capable of shaping a company's trajectory in a world where AI-driven innovations are reshaping industries at an unprecedented pace.

AI Infrastructure

In the dynamic landscape of Generative AI, a prominent challenge has come to the fore: the existing infrastructure is struggling to meet the surging compute demands of increasingly intricate models. As generative AI techniques, including GANs, transformers, and VAEs, evolve and diversify, their appetite for computational resources grows exponentially. This surge in complexity and scale places a significant strain on conventional AI tech stacks, resulting in a critical need for a comprehensive reassessment of AI infrastructure.

Organizations are now compelled to reevaluate and overhaul their AI tech stacks to unlock the full potential of generative AI. The conventional compute resources that sufficed in the past are now proving insufficient to handle the intricate calculations required by these models. This has prompted a shift towards innovative solutions such as leveraging cloud-based resources, harnessing the power of distributed computing, and investing in specialized hardware like GPUs and TPUs.

The revaluation of AI tech stacks isn't just an option; it's a necessity. The boundless applications of generative AI, from creative content generation to data synthesis, demand more than just incremental improvements in infrastructure. The journey towards more realistic, context-aware, and adaptive AI-generated content necessitates a resilient, adaptable, and high-performance tech stack. Thus, organizations are compelled to make strategic investments and adjustments to accommodate the escalating compute demands and to fully exploit the potential of generative AI in this new era of possibilities.

Regulation

Regulation of Generative AI varies globally, reflecting diverse approaches. The EU is pioneering comprehensive oversight through the AI Act, striving to balance innovation with ethical concerns. This proactive stance enforces stringent requirements, obliging businesses to adhere to transparent, accountable, and human-centric practices. In contrast, the US has opted for a lighter touch, favoring innovation by emphasizing existing frameworks and self-regulation. This approach aims to foster AI growth while relying on companies to self-monitor. The UK, however, takes a middle path. It promotes a risk-based model, prioritizing transparency and accountability. This balanced strategy seeks to offer regulatory clarity without stifling progress. As Generative AI advances, businesses worldwide face the challenge of navigating these distinct regulatory landscapes to harness its potential while upholding ethical standards.

Summary

In conclusion, while Generative AI holds immense promise, it comes with significant risks that cannot be overlooked. Cautious adoption and thoughtful risk management are essential for businesses to harness the true potential of Generative AI while safeguarding their interests and the interests of their customers. As the technology evolves and regulatory frameworks take shape, a more balanced and sustainable landscape for Generative AI may emerge. Until then, the path forward requires a delicate balance between innovation and prudence.

Generative AI has advanced rapidly in recent years, enabling exciting and diverse developments and applications across various industries. The last year especially has seen a big leap in the AI hype curve. From marketing, finance and education, to gaming, social media and even space travel, the possibilities of generative AI seem endless. Here, we explore some of the best publicly available use cases for generative AI, showcasing how the technology is transforming many different services and sectors.

Find out more about the big changes on the way at the Generative AI Week.

NASA unveils ‘alien’ AI-designed structures

The truth is out there – NASA has discovered alien technology. Well, in a sense.

NASA's Goddard Space Flight Center in Maryland has been utilizing commercially available AI software to design specialized components – known as "evolved structures" – for their missions. These structures possess a distinctive and unconventional appearance – or as NASA research engineer Ryan McClelland puts it: “They look somewhat alien and weird.”

However, it’s not just the way they look that sets them apart; according to NASA, the one-off parts weigh less, tolerate higher structural loads and are developed in a fraction of the time it takes humans to do the same work.

The design process begins by analyzing specific requirements of each mission. A computer-assisted design specialist maps out the connections between the component and the spacecraft or instrument, and then the AI software takes over, generating intricate and complex structural designs in short time frames.

While the algorithms play a significant role in the development, human oversight and intuition remains essential, says McClelland, who admits that the algorithm occasionally produces structures that are excessively thin.

However, the engineer emphasizes that analyses have demonstrated the algorithm-generated parts to have much lower stress concentrations compared to those produced by humans, contributing to a significant reduction in the risk of failure.

What’s more, the streamlined development process allows designers to allocate more time to other critical aspects of the mission.

The evolved structures have already found applications in several NASA missions, including the EXoplanet Climate Infrared TElescope (EXCITE) project, which employs a balloon-borne telescope for studying exoplanets with characteristics similar to hot Jupiters in orbit around distant stars.

Coca-Cola creation platform quenches thirst for AI

A Coca-Cola collaboration with OpenAI and Bain & Company has seen the development of an AI platform called ‘Create Real Magic’, which effectively opens up the Coca-Cola marketing archives to the public for the creation of original artwork.

This platform combines the capabilities of OpenAI’s GPT-4 for the generation of human-like text, and DALL-E for the creation of images based on text.

Digital creatives can produce a canvas for their AI-powered artistic experiments by using a variety of Coca-Cola branded elements – such as the famous script logo, the contour bottle and storied symbols from Coke advertising campaigns.

Artists have also been invited to enter their work into a contest, with the chance of having their art featured on Coca-Cola's digital billboards in Times Square, New York, and Piccadilly Circus, London.

“We will begin to leverage OpenAI’s technology in our marketing function to reimagine how we produce creative content, increasing the velocity from weeks to days,” says Manolo Arroyo, Coca-Cola’s global chief marketing officer.

“We see many applications of AI – including content creation and rapid iteration, hyper-personalizing content and messaging for consumers and customers, and driving two-way conversations with consumers,” he adds.

Coca-Cola is also exploring ways to leverage AI beyond marketing, in areas such as internal knowledge management and workflows, customer service and ordering, and point-of-sale material creation in collaboration.

Decades of data helps to create Bloomberg GPT

In March 2023, Bloomberg released a research paper on its development of a generative AI tool known as BloombergGPT.

Trained on Bloomberg’s massive datasets, BloombergGPT is a large language model (LLM) developed to support a range of natural language processing (NLP) research and analysis tasks within the financial industry, including sentiment analysis, named entity recognition, news classification and question answering. It also enables Bloomberg to harness the huge amount of data available on the Bloomberg Terminal, which has been the go-to financial market data resource for traders over the last 40 years or more.

The company’s machine learning (ML) product and research group collaborated with the AI engineering team to combine financial language documents from Bloomberg's extensive archive with public data. The resulting training dataset was used to develop the 50-billion parameter decoder-only causal language model.

According to Bloomberg, the model outperforms existing open models of similar size on financial tasks while performing on par or better on general NLP benchmarks.

“The quality of machine learning and NLP models comes down to the data you put into them,” says Gideon Mann, head of Bloomberg’s ML product and research team.

“Thanks to the collection of financial documents Bloomberg has curated over four decades, we were able to carefully create a large and clean, domain-specific dataset to train an LLM that is best suited for financial use cases.”

Microsoft’s Bing turns to ChatGPT to find an edge

In the battle for search engine dominance, Google has been the historic victor, but earlier this year Microsoft unleashed a weapon to mount a serious fightback.

In February 2023, the tech giant unveiled its AI-powered Bing search engine and Edge browser. The AI technology comes courtesy of ChatGPT – which is not surprising considering that Microsoft invested an estimated $10bn in OpenAI, ChatGPT’s developer, in January.

However, there are differences between using ChatGPT – which was released to the public in November 2022 – and AI-powered Bing. For example, the publicly available version of ChatGPT works on GPT-3.5 and is trained on web data from before 2021, while Bing AI works using GPT-4 and, of course, references the current web in its entirety. Also, Microsoft says its proprietary Prometheus model optimizes the OpenAI model, and claims that this results in more relevant and targeted search results.

Moreover, Microsoft’s aim is to offer an integrated experience that combines search, browsing and chat functionalities.

The enhanced Bing search engine incorporates an interactive chat, allowing users to refine their searches by asking for more details. In addition, the new Bing can generate content, helping with tasks like writing emails, creating travel itineraries or preparing for job interviews.

The Edge browser, meanwhile, has been updated with AI capabilities, introducing features like Chat and Compose. Users can request summaries of lengthy reports or ask for assistance in composing content for social posts, for example.

As you might expect, Google isn’t a mere spectator when it comes to generative AI and is investing heavily in its own development, known as Bard.

LinkedIn opens up to OpenAI

Another Microsoft-owned platform, LinkedIn, is also taking advantage of advanced OpenAI GPT models, including GPT-4, to enhance user experience. Tools offer a range of features, such as generative AI-powered collaborative articles, personalized writing suggestions, and AI-generated marketing copy.

The collaborative articles feature begins with AI-powered conversation starters crafted by LinkedIn's editorial team. Leveraging the platform's Skills Graph, each article is then matched with relevant member experts who can contribute their lessons, anecdotes and advice based on their professional experience.

Another feature, personalized writing suggestions, allows users to receive tailored copy advice based on their existing content. By analyzing the user's skills and experiences, the AI tool suggests the most important information to highlight in the ‘about’ and ‘headline’ sections of their profile.

The latest feature, announced in June 2023, offers AI-generated copy with the aim of increasing marketing efficiency. These suggestions utilize data from users' LinkedIn Page and Campaign Manager settings to provide five ad headlines and copy recommendations, which the user can then select from and tweak.

Initially, the new feature is being tested in English with a select group of customers in North America. However, LinkedIn has plans to expand its functionality, supporting additional languages and making it available to a wider user base.

Writer to the rescue for HubSpot designers

When inbound sales and marketing software developer HubSpot wanted to enable its product and content designers to hit new heights with their UX writing, the company turned to Writer – an LLM-based generative AI platform that uses ML and NLP to tackle a range of writing tasks.

HubSpot’s aim was to make UX writing more consistent across the company’s product, as well as increase quality and speed up processes.

“We wanted a centralized and structured database of terminology that was important to us, so that we could work with our localization team, because defining terms in a more standardized way helps enable the best and highest quality translation,” explains Jonathon Colman, HubSpot senior design manager and content design practice co-lead.

According to Colman, before the partnership with Writer, HubSpot relied on “just a heap of disconnected spreadsheets and documents where we had defined terms and styles”. But now, the AI platform has allowed everything to be centralized in one place.

“Writer has taken what was this really chaotic landscape of just disconnected, disorganized things and made it all systematic. So everyone can move a lot faster, a lot better together.”

While Writer has improved organisation and the quality of writing, the HubSpot product designers are also now more confident in approaching UX writing tasks, Colman adds. What's more, time savings have allowed them to redirect their efforts towards more strategic work, such as customer journey mapping, improving accessibility and new product launches.

“That’s a much better use of their time than going into a wiki or a static document somewhere and rifling through a bunch of standards,” says Colman.

Expedia travels down the ChatGPT path

Travel giant Expedia is another company linking up with OpenAI’s ChatGPT to enhance customer experience.

April 2023 saw the beta launch of the Expedia travel app, allowing users to engage in open-ended conversations and receive recommendations for travel destinations, accommodation, transport and activities.

The intelligent shopping feature automatically saves hotels discussed in the conversation to a ‘trip’ in the app. The feature is designed to help users organize effectively when selecting dates, checking availability and adding flights, cars or activities.

Expedia has already leveraged AI and machine learning (ML) across its platform with personalized trip options generated using algorithms that consider various variables, such as hotel location, room type, date ranges and price points. In addition, Expedia's price tracking feature utilizes comprehensive flight data, AI and ML to compare current flight prices with historical trends.

An AI-powered virtual agent is also available 24/7 to assist travelers with unexpected trip changes.

"By integrating ChatGPT into the Expedia app and combining it with our other AI-based shopping capabilities, like hotel comparison, price tracking for flights and trip collaboration tools, we can now offer travelers an even more intuitive way to build their perfect trip,” says Peter Kern, vice chairman and CEO of the Expedia Group. According to the company, the beta testing phase will allow it to refine the experience based on user feedback.

Kern told CBS News: "We built our own AI to basically monitor the outcomes for what ChatGPT comes back with because, really, we only want to help people shop for travel. We are really using our own capabilities to monitor the outcomes, make sure travelers don't get strange responses. And if something goes wrong, we're trying to make sure it comes back to travel."

A new kind of language model for Duolingo

Language learning platform Duolingo recently collaborated with OpenAI to integrate GPT-4 into a new subscription tier called Duolingo Max. The upgraded tier provides learners with two new features: Explain My Answer and Roleplay.

Explain My Answer allows learners to engage in a chat with ‘Duo’, the Duolingo mascot, in which they can receive explanations on why their answers were correct or incorrect, seek further clarification and request examples.

Roleplay, meanwhile, enables learners to practice conversations in real-world scenarios with virtual characters. Following the conversation, Duo can give learners feedback regarding the accuracy and complexity of their responses, as well as offer tips for future conversations.

The initial launch of Duolingo Max covered Spanish and French courses for English speakers on iOS devices, with availability in the US, Great Britain, Ireland, Canada, Australia and New Zealand.

“Artificial intelligence has always been a huge part of our strategy,” says principal product manager Edwin Bodge. “We had been using it for personalizing lessons and running Duolingo English tests. But there were gaps in a learner’s journey that we wanted to fill: conversation practice, and contextual feedback on mistakes.”

Robolox building towards a generative AI ecosystem

Roblox, the game platform known for its immersive 3D experiences, has signaled its intention to “revolutionize creation” through generative AI.

Currently, creators on Roblox use Roblox Studio, a free platform, to build their experiences and publish them across various platforms. However, Roblox says it is building a platform utilizing AI tools that will “enable every user to be a creator – not just those comfortable with Roblox Studio and other 3D content creation tools”.

Roblox plans to introduce tests for two new tools in the near future: generative AI text prompts and generative AI code completion. According to the plans, the tools will enable creators to develop integrated 3D objects with built-in behavior. For example, a creator will be able to design a car by describing its attributes, and generative AI will do the rest.

The ambition is for tools utilizing voice, touch-based gestures and other intuitive interfaces to be available to users in the future. Roblox also wants the AI community to contribute; the ambition is to foster an “ecosystem” where third-party AI creation services integrate directly into the platform.

When announcing the plans, Roblox acknowledged the importance of implementing generative AI “thoughtfully and ethically”. The company also emphasized that safety, moderation and the economic aspects of the platform would need to be carefully addressed.

Find out more about the big changes on the way at the Generative AI Week.

Interview with Khalil Maaouni, Head of Data and Digital at Coca-Cola Bottlers Japan.

There is a worldwide wave of excitement around the revolutionary potential of generative AI.

On October 16th-19th 2023, that global buzz will be concentrated in Atlanta, Georgia, where experts from around the world will congregate for Generative AI Week, with topics under discussion including data leverage and balancing capability and risk.

One of Generative AI Week’s expert speakers – Khalil Maaouni, head of data and digital at Coca-Cola Bottlers Japan – is at the cutting edge of both those topics. He believes that one of generative AI’s key attributes is its potential to “dynamize access to data”. This involves making data “a lot more available faster” as well as “getting a pulse on what is being looked at in the company in terms of the context of data”.

According to Maaouni, this ability to explore the context of data means generative AI can fill important gaps in companies’ current capabilities.

Download interview here >>>

Interview with Daniel Hulme, Chief AI Officer at WPP. 

Organizations spanning a wide range of sectors are beginning to recognize the boundless opportunities and uses for generative AI in gaining a competitive edge. These possibilities are set to be explored in detail at Generative AI Week, taking place in Atlanta, Georgia, on October 16th-19th 2023.

Among the world’s foremost AI experts speaking at the event will be Daniel Hulme, Chief AI Officer at WPP. Hulme has a quarter of a century’s worth of experience in AI within academia. He is also entrepreneur in residence at University College London and has acted as an expert witness to the UK all-party parliamentary group on artificial intelligence. So when he talks about generative AI, it makes sense to listen. And significantly, he’s never been more excited about the power and potential of the technology than right now.

Download the interview here >>>

While most enterprise businesses are running Generative AI pilot projects, typically they are taking a cautious approach to customer facing applications due to the reputational, financial and brand implications should something go wrong.

Booking.com are an example of a business that have already rolled out a customer facing product with Generative AI at its core, the AI Trip Planner. The planner incorporates the benefits of LLM’s with the value of their internal data, to deliver a differentiated product in the travel market.

In this interview with Charlotte Munro, Global Product Marketing – Generative AI at Booking.com she covers the business case behind the product, how they build the internal team that worked on it, the risks they faced and how they managed them, and the early feedback from their customers.

Why did Booking.com decide to integrate Generative AI into its products?

• While the AI Trip Planner is a new product that we developed and launched in July of this year, AI and ML has always been an integral part of Booking.com, recommending destinations and options to millions of travellers every day on the platform. We have been using AI extensively for years in order to optimise interactions with our customers who are both travellers and partners - from laser focussing / contextualising / tailoring recommendations and the user experience, through analysing the content and pictures provided by our partners and customers. It is the strong foundations of AI within Booking.com’s ecosystem that allowed us to develop a robust AI Trip Planner so quickly.
• We believed then, and we continue to believe now, that we can build a more compelling and differentiated offering if we can leverage AI technology to deliver a more tailored and relevant booking experience, a connected trip, that would be more responsive to a booker's needs and help manage the different aspects/components of their trips.
• While we make some use of OpenAI’s ChatGPT to partially power the new conversational experience, Booking.com’s existing machine learning models enrich the information presented to customers with tailored destinations and accommodation options.

How did you assemble the team to create your first customer facing Generative AI products? What job functions were represented?

• We assembled a virtual task force team of colleagues that represented all functions of our business. This included our ML experts, Product, UX and UI as well as our research, product marketing, legal and PR teams.
• To compliment the technical skills and requirements, and ensure we created a product that resonates seamlessly with our customers, we then focused on ensuring we had the right customer insights, legal protections, and impactful key messaging. This meant we not only had a great MVP product, but also one which should have longevity build in
  

How are you ensuring that customers get a differentiated experience when using Generative AI?

• We know Booking.com is really good at helping our users book their trip, especially when they know where they want to go and have a fair idea of when they want to travel. However, we know that before travellers get to that point there is a lot of discovery and inspiration that happens. The AITP allows users to start their search and discovery based on their trip intent - whether it be a beach holiday with their family or a hiking adventure with their friends or simply exploring a range of potential destinations and options. This means we are interacting with customers throughout their entire trip journey – from ideation through to bookings.
• The AITP allows for a more unconstrained search experience where users can quite simply tell us what they want, and we show them options that suit their needs. Travellers can ask the AI Trip Planner general travel-related questions, as well as more specific queries to support any stage of their trip planning process. This includes scoping out potential destinations and accommodation options, providing travel inspiration based on the individual traveller’s needs and requirements, as well as creating itineraries for a particular city, country or region.
• The AITP also really helps travellers by saving them time and effort - doing the hard work for them.
• We are always looking for new ways to make customer interaction smoother and richer and our recommendations have a seamless visual UX which really helps our travellers explore and get an idea of a potential destination or accommodation for their next trip.

How did you manage the risks associated with the technology, such as privacy and security?

• Our customers' rights, privacy and security is always of the utmost importance to us, which is why we engaged our legal experts from day one of the build. The AI Trip Planner went through a robust product review process and has been developed according to rigorous standards to ensure a safe and inclusive environment for travel exploration. We used the same robust legal protections that we use with all of our products and this includes custom AI moderation layers for the AITP, which detect and block harmful content, and ensure our customer’s data remains safe and compliant with all regulations.
• We also carry out robust data protection impact assessments to identify any areas of potential risk and to ensure compliance with the applicable data protection laws
• With all our travel experiences we want to ensure that travel is a safe and inclusive environment for everyone, and this traveller experience carries through to how customers interact with the AI Trip Planner.

How do you protect personal information and prevent hallucinations?

• Booking.com has built a custom moderation layer to structure the collaboration between the LLM and Booking.com’s machine learning models, as well as guardrails to detect and remove certain personal information and/or non-travel related content. This also works together with OpenAI’s moderation to help identify and block potentially harmful content. The aim is to remove personal data when it’s unnecessary to the search and travel planning process.
• To minimise and mitigate any inappropriate and inaccurate responses being surfaced to travellers, we have introduced moderation content filters which remove any offensive questions and informs the user very clearly that the tool cannot and will not respond to such questions.
• We have made some good strides in moderating and filtering out inappropriate content and will continue to innovate and develop within this space to optimise further and incorporate learnings to ensure accuracy and relevancy of the content that is surfaced. We have also carried out robust data protection impact assessments to identify any areas of potential risk and to ensure compliance with the applicable data protection laws.

How do the LLM and Booking.com work together to answer a question?

• The AI Trip Planner is a combination of third-party LLM technology and Booking.com’s own machine learning recommendation systems. The LLM technology that we are currently leveraging from OpenAI enables the AI Trip Planner to quickly read and understand questions that are being posed in the tool, as well as to provide responses in a conversational, natural way. The LLM also references a wealth of other data to provide certain kinds of recommendations, for example itineraries with suggestions of great things to do in a specific destination.
• Where the Booking.com machine learning models take over are when the data needed to respond to the question is fully in Booking.com’s realm of expertise, for example with specific destination and property recommendations. All of the visual elements in the tool, including pricing information are from Booking.com.
• To make it even more concrete/simple, conversational text and answers to general travel questions are provided by ChatGPT. For example, what’s the weather like in Bali in December? Or what’s a good itinerary to visit the north of Italy for a week? The visual links for specific destinations and property recommendations, as well as the moderation layers that help keep content travel-related and safe come from Booking.com.

Do you have any initial market feedback on how the product has been received?

• The objective of rolling out the MVP slowly and market by market allows us to continue to develop and optimise the product based on how our customers are interacting with the product and what they are using it for/type of broad queries that they are asking. We are seeing travellers explore popular destinations as well as themes like beaches and family trips. Itineraries are also popular. What would you do differently if you ran the project again?

What do you wish you had known before developing this product?

• I think the nature of AI and this type of emerging technology means that you are never going to have the perfect test and launch conditions. Which is why developing an MVP that we gradually roll out to our customers and new markets, allows us to improve the product and the experience. As we can get early signals from our customers - how are they interacting with the product, what kind of themes they are exploring and how are they using it? We will continue to evolve the AITP and ensure that the product meets the needs of our customers.
  

The media crescendo surrounding generative AI has reached such a pitch in recent months that it’s difficult to ascertain genuine insight from all the noise. This means it’s more important than ever to go directly to the experts working at the cutting edge to discover the true significance of the latest developments.

Well, the expert opinion is in: believe the hype, because “there’s a big change coming”.

That’s the view of Chris Booth, generative and conversational AI consultant and product owner for NatWest Group’s artificially intelligent agent: Cora.

So where are we now with generative AI, and where are we heading? When assessing the potential impact of generative AI and the disruption that could be coming down the line, Booth says it’s helpful to think in terms of where we are on the ‘sigma curve’.

“What I mean by that is, if we’re at the top, then most of the impact has already happened and we won’t see much change going forward,” he explains. “If we’re in the middle of the curve, then we can still expect to see generative AI applied to other technologies in the future. Or are we at the start of the sigma curve, with big changes to come?

“Overall,” he says, “that’s where I am – I think there’s a big change coming.”

Booth is one of the experts speaking at the Generative AI Summit, taking place at Hilton Syon Park, London, on 16th and 17th May 2023. He’ll be addressing the topic of what generative AI means for chatbots, drawing on his direct experience of working with NatWest’s Cora chatbot.

Powered by IBM Watson, Cora operates in the closed domain, which is where chatbots primarily exist – especially in large organisations – responding to action- or task-oriented questions such as, ‘Can you change my address?’

“It’s basically a large logic tree,” Booth explains. “This means we dictate what buttons are presented and then what button you click obviously changes your path. So that’s closed domain, and it works really well.”

However, closed domain and logic trees can have limitations, says Booth. “Multiple trees are brittle. They can become very difficult to manage and maintain as they grow at an exponential rate. And the larger the tree, the more links you have to manage and it’s a mess.”

But this is where generative AI has the potential to change things massively, Booth insists. While he admits that generative AI is “nothing new” in the natural language processing (NLP) space – and that it’s only in the last few years that models have become good enough to make generative AI “a contender” – Booth is excited by the potential for changes it could bring for the “opposite end” – the open domain.

He explains: “The open domain deals with questions like, ‘What did Obama do before he was president?’ – an open-ended question that can be difficult to answer. And that’s where logic trees really struggle to capture the potential scope and possibilities of how you can answer that question.

“So that’s where generative AI has huge potential for expansion, with the potential of opening new use cases for businesses to approach.”

However, there are challenges and possible drawbacks. Among them are transparency and explainability.

“Generative AI is usually powered by language models – deep AI machine learning,” Booth explains. “And these deep neural networks have billions and billions of parameters, which makes it difficult to distinguish and understand how the AI has come to its decisioning.”

Also, the language models can be prone to ‘hallucinations’ – which Booth describes as “a fancy word for outputting nonsensical and incorrect answers”. From a language model perspective, these can be very difficult to control, he says. And added to these issues are obstacles surrounding cost, privacy and data security.

But despite the challenges, Booth believes everyone will be putting generative AI to practical use at some point. “There’s going to be varying degrees of how quickly it happens. There are already plenty of startups based on ChatGPT and GPT-3. And there are small businesses in marketing, for example, that are going absolutely nuts with ways of slowly automating things.”

What’s more, opportunity is ripe for breakthroughs in the development of generative AI. “There’s the potential to make a massive impact,” Booth insists. And he’s hoping to realise that potential himself. He reveals: “I’ve got a project right now I can’t talk about in detail, but we’re trying to find ways to cover the gaps and weaknesses of language models. If we do, the implications are pretty large.”

Find out more about the big changes on the way at the Generative AI Summit.

Given the legal challenges that have already arisen as companies dispute both copyright and legal ownership of the output of Generative AI models, this session will cover:

• The existing legal framework and its limitations when addressing Generative AI
• Copyright risks – what are the current risks and how can they be mitigated
• Understanding ownership – who ultimately owns the output from Generative AI

Watch back Matt Hervey, Head of Artificial Intelligence Law Gowling WLG video presentation here >>> 

Chris’ role at NatWest involves him leveraging Machine Learning to improve the effectiveness of their conversational AI agent ‘Cora’ which currently involves investigating how to integrate the benefits of LLMs (Large Language Models) like ChatGPT without any of the downside risk. 

In this session, Chris cover:

• Assessing the ability of LLMs to generate more human sounding output due to the sheer size of the dataset they have been trained on
• Understanding the limitations of LLMs for the enterprise applications due to the risks associated with incorrect or misleading answers
• Blending LLMs with additional architecture, such as knowledge graphs, to extract the benefits while minimising the risks

Watch Chris,Product Owner - Machine learning and Insights at NatWest presentation here >>>

Generative AI represents a significant advancement in the field of Artificial Intelligence, offering new opportunities for creating content and data, improving decision-making, and automating tasks. This presentation will explore how it can change the way we interact with technology. 

Watch Nina, author and Generative AI expert presentation here >>>