- il y a 2 jours
AI and The Future of Commerce
Catégorie
🤖
TechnologieTranscription
00:00So, good afternoon, I'm Jennifer Schenker, Editor-in-Chief of Innovator, a publication about the global tech industry, and it's
00:10my great pleasure to welcome here today Rajat Taneja, President of Technology for Visa, and we're here to talk about
00:19AI and the future of commerce.
00:23So, Rajat, tell us, how do you think AI is going to change commerce?
00:30You know, AI has been the topic of this conference, it's been the topic of conversation for the last 18
00:36months, and it's exciting because generative AI is going to have a profound impact on society, and we are seeing
00:44the green shoots of that already, and it will have a big impact on commerce.
00:50I think it will be in two areas.
00:52The first one is the shopping experience itself.
00:56The last big inflection point in shopping was with the internet, which gave birth to e-commerce and omni-channel,
01:05and brought in a lot of goodness in connecting consumers and merchants cross-border, breaking barriers of time and distance.
01:15But if you think about it, shopping is still fraught with friction, it is a hunt and peck, you are
01:22looking for things on a search engine, on a platform, and it's not the personalized experience that we have dreamt
01:30of for the longest time, and the impediment has been technology.
01:33And with generative AI, you can imagine a world where you have a model, that's your digital twin, that is
01:42able to have an experience for you that's end-to-end, where you can purchase and plan and manage what
01:52you are buying seamlessly, and everything is hyper-personalized.
01:57It's like self-driving of shopping, and it allows everything to be tailored and relevant to you as a person.
02:07So that's the exciting part.
02:09There is another part of it, which is a bit scary, which is the fraud and the cybercrime that comes
02:17with the territory.
02:18In 2022, the cost of cybercrime globally is estimated to be $7 trillion, and it will be $10 trillion in
02:302025, growing at a 32% growth rate compared to global economies that are growing at 5%.
02:37So, the bad actors and hackers are as excited by large language models as we are, they are having posts
02:47and conversations in the dark web preceding the conversations we are having.
02:53We have ChatGPT, they have FraudGPT, and WormGPT, and DarkBert, and that is allowing them to create campaigns of social
03:03engineering and deepfakes,
03:06and try to defraud consumers in a way that was never possible before.
03:12So, the other part of generative AI is a responsibility we have as platforms to create protection and security and
03:21use the models to stop attacks,
03:25fraudulent attacks like enumeration attacks, when hackers are taking credit cards and seeing which ones work, which ones don't, and
03:32quickly cashing it out.
03:33So, how do we manage and build those large risk models to combat it at the very inception?
03:42Last year, we stopped $40 billion of fraud.
03:45Wow.
03:46We are attacked over 90 million times on our APIs every month.
03:5111 million emails trying to inject malware.
03:55So, it's a very dangerous world, and it's going to be an arms race on building models and protections for
04:01both cyber and fraud.
04:02And so, Visa is really trying to play in both areas.
04:07You want to offer customers the advantages of Gen AI on end services.
04:12You're also helping to fight cyber crime.
04:15And I think that's important to note that although many people think of you as a credit card company, you're
04:21actually a technology platform.
04:23So, and you've spent $3 billion over the last 10 years to implement AI in all parts of your business.
04:32So, talk to us about, you know, give us a holistic view of what that platform looks like today and
04:39where Gen AI is going to take you.
04:41Jennifer, that's absolutely correct.
04:43Visa is the world's largest SaaS platform, actually.
04:46It was a SaaS platform before the word SaaS even existed.
04:51We process 290 billion transactions in real time every year.
04:59We run $15 trillion through our network.
05:03We have a private network that connects our financial institutions, merchants to us.
05:09That's 20 million route miles long, one of the biggest networks.
05:14And so, it all begins with a payload of data.
05:19When you're buying a cup of coffee and you tap your card or you swipe your card, a payload of
05:24information and data comes to us.
05:26And that begins the journey through the ecosystem of us enabling you to get the coffee or buy the subway
05:33ticket or transact for any goods or service.
05:37And so, the data is the most important asset, not just at Visa but at every company, actually.
05:43And over time, data gets siloed.
05:45Structured data and unstructured data don't talk to each other.
05:49You have documents and spreadsheets and other contracts and other content.
05:55You have enterprise data warehouses.
05:58The data is highly transformed.
06:00So, the investments that we made and I think is imperative to really harness the power of large language models
06:08is your data.
06:10In any AI, there are three elements.
06:13It's the compute, it's the model and the data.
06:15And the data is what will distinguish and be the secret sauce for any company out there.
06:20So, we stepped back 10 years ago and rebuilt our data platform from scratch to try to lean into a
06:30world where deep learning and predictive AI was very important for serving our clients, managing risk, managing fraud, managing cybersecurity.
06:39And now that is enabling us to lean into generative AI in a big way and serve the ecosystem.
06:46So, you've talked a little bit about how Visa is helping protect against cyber attacks.
06:55What are you doing to protect consumers' data and data privacy and control over their data?
07:03Yeah, I think that's a very important obligation we feel and we are wanting to help with because consent and
07:11transparency and privacy and compliance is at the very heart of Visa.
07:16And just like we brought that to payments, our goal, our hope is to bring that to data and to
07:24see what we can do to contribute that scale to enabling merchants and consumers to interact with each other where
07:33the data is private.
07:35It is secure.
07:36It is secure.
07:37It is never provided without consent.
07:40And can we help build the digital twin of the consumer with features of a model that can be provided
07:48by them to a merchant to personalize their experience?
07:52So, this concept of data tokens is what we are thinking about analogous to what we do with payment tokens
08:01where your credit card or debit card is tokenized when you have it on your phone and it protects you.
08:06So, it's an exciting concept but it comes with a huge obligation of doing it correctly with security, compliance and
08:14control and consent.
08:15Okay, thanks.
08:17So, now we have a better idea of what Gen AI will allow you to do for your customers.
08:24How is Gen AI changing Visa internally?
08:30How many of your employees are using it?
08:34What sort of training are you doing?
08:36How is it changing or improving your productivity?
08:40You know, given our heritage with technology and with AI, our first model went live in 1993.
08:48We had been working with transformers and with large language models well in advance of when GPT 4.0 and
08:56ChatGPT was released.
08:58But we were also incredibly impressed and marveled at the power of that large language model and what it could
09:06do.
09:07So, we leaned in right at the beginning, realizing that it could have a profound impact not just in what
09:14we could do for our clients,
09:16but also internally.
09:18So, we have provided a secure access to GPT to all employees.
09:24And we have connected it to our own data so that we can empower our employees with a lot more
09:31insight
09:32and help them with their day-to-day jobs and reduce the friction.
09:36But we do it by maintaining that data in our premises and keeping it very safely.
09:41All our engineers have pair programming available to them through Copilot so that when they are coding,
09:50you have an assistant helping you write code, help you with quality and bugs and testing.
09:57We are doing the same thing with automating our business functions like customer support, internal support.
10:03We are trying to open it up in our sales functions, marketing and finance.
10:09The biggest area was risk, managing risk, ecosystem and payment fraud disruption.
10:14So, we have leaned in and we are noticing a very positive impact and it's only the early days.
10:21Okay.
10:23So, getting back to cyber security for a moment, cyber attacks and deep fakes are not the only worry.
10:34Encryption threatens to be broken soon and financial services obviously are going to be the first and potentially the worst
10:47hit by this.
10:48How is Visa preparing for this post-quantum era?
10:53Jennifer, that's a great question.
10:55If you fast forward maybe 20 years from now or maybe 15 years from now,
10:59the next exponential technology wave will be quantum computing.
11:04And quantum computing has capabilities that will have a dramatic impact on society
11:12because unlike classical computers which work on information at zeros and ones, on and off,
11:19and that's how it processes and provides its computational capability,
11:25quantum computing uses the physics, quantum physics and quantum mechanics
11:31where you don't have zeros and ones, you have qubits, quantum bits,
11:36that can exist both as zero and one simultaneously so that increases the addressable space
11:42and the computational power that is going to be available.
11:46But it will take a lot of time and effort to create that type of large high-scale infrastructure.
11:52But when it is available, on the negative side of what it can do is it will break the encryption
11:58that is used predominantly on the internet for transmitting secure transmissions,
12:04public key cryptography or asymmetric encryption will get broken,
12:09which so far classical computers are not able to do.
12:12So companies like ours feel a real obligation to be working on this
12:16and for the last 10 years we have had a big effort on this to think about post-quantum cryptography.
12:22We have done a lot of research, we have submitted papers,
12:25we are working with standards bodies like the U.S. NIST and EMVCo
12:29to envision a world where the next generation of cryptography will not succumb
12:35to quantum breaking RSA and ECC protocols.
12:39Okay, thanks. And so last week, jumping to another topic here,
12:45the European Central Bank said publicly that the use of AI in finance
12:51needs to be monitored and possibly regulated to prevent harm to consumers
12:56and ensure the proper functioning of markets.
12:59And while the ECB recognized how Gen-AI could help in cyber crime fighting it,
13:07it also warned of risks such as hurting behavior, over-reliance on a limited number of providers
13:14and more sophisticated cyber attacks.
13:17What, in your opinion, is the best way to ensure that we use AI responsibly?
13:24I think that's incredibly important.
13:26We talked about the science of AI, the compute, the data, the models,
13:30but there is an art to AI as well.
13:33It's the people, it's the policies, it's the governance and the oversight.
13:37And it is imperative upon practitioners as well as policymakers to really think about responsible
13:44and ethical use of data and AI in the models so that they work for the best interests of society.
13:52So, for example, at Visa, we have a data use policy that is documenting our use of data for models.
14:00And when a model is created, another independent team tests it, and then there is a model risk management team
14:06that governs all of this.
14:08And we put it into production, we oversee it for drift and behavior so it doesn't have biases, hallucinations, etc.
14:15So that's very important.
14:17And governments and regulators like in the EU and in the US with the executive order there last year
14:24and the EU AI ruling that's going live and Singapore's AI verified toolkit are incredibly important.
14:32I think the best policies, however, will require a collaboration between regulators and government,
14:38academia and industry together.
14:42And because AI doesn't work in the jurisdiction of countries and it works globally,
14:48it's important for all of this to have a global perspective like the conversations taking place this week in South
14:55Korea
14:55in the safety consortium on bringing in all these people together.
15:01I think that's really important.
15:02We have signed in for this in letter and spirit and we believe in this deeply.
15:07Okay, thanks.
15:09So now I want to ask you, you know, Visa has been using AI in some form for over 30
15:16years,
15:16but many large companies here in Europe and even elsewhere in the world are just getting started
15:24and many are still at the POC stage.
15:27So what advice would you give them about how to scale AI and use it in a way that really
15:34moves the needle for their company?
15:36My first piece of advice is don't panic.
15:40In the big arc of time, multi-decades of generative AI, this is still day zero.
15:48And there is a lot that's going to happen and it sometimes might feel like everybody's doing something faster than
15:54you.
15:55But I think it's important first and foremost to really understand what these models are, what they can do and
16:01really deeply understand how they are created
16:04and how you can supplement them with your own magic and your own secret sauce, which is your data.
16:10And then you should have a written strategy.
16:12How do you want to use it internally for productivity, for business automation?
16:17And then how do you want to use it for your clients and your customers?
16:21You then need to sit back and look at your data infrastructure and say, are you geared to really take
16:27advantage of this
16:28and what investments you need to make in your data infrastructure, the way you ingest data, the way you store
16:33data,
16:34the way you process data, the way you connect it across your company.
16:38And then you need to put in place policies upfront that reflect your values in how you will govern and
16:46manage responsibly and ethically data
16:49so that it is compliant and has a great focus on usage.
16:56So Visa, now we're going to turn to the startups in the audience.
17:00Visa has a new $100 million Gen A fund that will invest in the next generation of companies.
17:09Tell us a little bit about this fund and how you want to engage with startups in this room and
17:14elsewhere.
17:15Well, I'm so happy and pleased to be in a company where this is so important that we have a
17:20$100 million focused fund for Generative AI.
17:25And we have a very active venture arm, but this fund is specifically to help accelerate Generative AI applications
17:35and models that will do good in society.
17:39And if you think about it in concentric circles, the closest circle is how it can impact commerce and payments
17:46in a very positive way.
17:48But then you can go to adjacencies and see how it will help with cybersecurity or operations
17:54or how it will help with business automation, how it will help with productivity.
17:59So it's concentric circles and we want to encourage, engage and have conversations.
18:05We are having active conversations with over 300 companies right now.
18:09300 startups.
18:10And we would welcome you if you have an interest and are working on Generative AI to come in and
18:15talk to us and have a dialogue with us.
18:19You can also use our infrastructure and our scale on risk and other capabilities through APIs that we have exposed
18:27for almost everything we do through our developer platform, which is developer.visa.com.
18:33So between the developer platform and the Generative AI fund, I hope we can catalyze even more thinking, even more
18:41excitement in this wonderful space.
18:43And so, you know, I guess we could sum up by saying that you are encouraging the people in the
18:51room to help build the future of e-commerce along with you.
18:56Let's build it together.
18:58And with that, I'm going to say thank you very much.
19:01Thank you.
19:01Let's have a nice round of applause for Rajat.
19:05Thank you.
Commentaires