Regulating AI and Twitter M&A

From Mahart.

We're Innovation, Money and Power Collie in Silicon Vallet NBN.

This is Bloomberg Technology with Caroline Hyde and Ed Ludlow.

I'm Carolin Hyde at Bloomberg's world headquarters in New York.

And I met Lovelow in San Francisco. This is Bloomberg Technology coming up.

The race to regulate artificial intelligence. It's underway in Washington as his Open AI CEO Sam Altman lays out the benefits and the risks to senators, and.

The CEO of Andresen backed Hippocratic AI joins us in studio to discuss their generative AI healthcare tech.

An X, the parent company of Twitter, has made as first acquisition a tech talent recruiting service called Laski. We discussed why verse four almostlet's get to these markets because once again we're actually seeing big tech on top lackluster trading day muted as we worry about the debt ceiling. The arguments still abound between left and right. We're seeing that's that one hundred and five tens percent big tech

Right now.

Across the world, we're seeing tenure yield actually up six bass points, even as we see in the individual retail data coming out showing you and I are still willing to spend in this inflationary data.

It seems as though.

Some of that federal reserves speak coming from Loretta Mesta from Barkin seem to be hinting the look that they're still all.

Eyes on inflation.

Let's someone look at what's happening in terms of bitcoin as well. Because the dollar is out performing, that means bitcoin's on the downside. We're off by about a percentage point in the day, still languishing around twenty seven thousand. ED dig into some of the big movers because there are points in the up side from big tech today.

It's like a market that's treading water when you think about the macro, but when you think about the micro, there are some stories out there Tesla up one point six percent the AGM after the bell today, but Bloomberg reporting overnight, according to sources, Shanghai is moving towards trial production of an updated Model three that's supporting the stock. Baidu, speaking of China, also up three point three percent. Strong earnings,

And how to regulate it.

And right now Sam Altman, the CEO of Open ai, really telling law makers on Capitol Hill about perhaps steps necessary to put in place rules around AI technology. He says is so powerful that he's worried it could have repercussions quote an a level far beyond anything we're prepared for. He also said, look, it's going to choose the labor market.

Take a listen, GBT four A will I think entirely ontomate away some jobs and it will create new ones that we believe will be much better.

Pleased to say, we've just pulled out of that hearing ed Anna Edgerton, who's been really listening into not just Sam Altman, but we've also heard from representatives from IBM who, well we know that IBM is already saying they're cutting their back office staff most likely calls of AI and scientists represent in too.

What's the key takeaway thus far?

It really has been all focused on regulation. You have Sam Altman of open AI and Christina Montgomery, the chief Privacy and Trust Officer for IBM, almost pleading with lawmakers to regulate this space. They want guardrails and they want certainty to protect against the most dangerous abuses of this powerful technology. And we hear it from the senators themselves.

While you join us from the hill and some outman continues to speak, and he's saying that the US should form an agency to license some AI efforts. A lot of bench capitalists and founders I know do not agree with the idea of licensing. They think it's basically a block to innovation. When you have hearings like this, also depends on the questions that get asked. Sometimes these lawmakers

There have been some thoughtful questions on this and He's had some thoughtful answers as well. He made a really interesting point when he said, you know, the US can leverage its leadership in the development of microelectronics and ships that are needed to run.

These powerful AI systems.

The US can leverage that leadership to encourage global governance so that not only are we setting rules here in the United States, but he made the point that we're not the only ones developing this technology. This is something that has a potential to impact the whole world, and for when it comes to governance of this new technology, it really needs to happen on a global level.

To that end, many will be unsurprised to know that the EU has kind of been leading the regulatory charge. They're already proposing potential laws in the next month that'll be eyed by.

The entire parity of the European Commission.

I'm interested in how much you think we'll make is all seeing what Europe is looking at all thinking about the global ability to regulate.

Well, certainly US lawmakers are casting and their counterparts over in Europe. And Christina Montgomery of IBM had an interesting little comment in her opening statement. She said, there's been a lot of hype around generative AI, but that doesn't mean we should move away from the risk based approach

Yeah, it really is about application for many at the moment, and Juton, we thank you so much to get her back onto that hearing at the moment and fantastic analysis coming from Capitol Hill. What about those that used to advise governments, what about those that are thinking about the application thus far of regulation. Lindsay Gorman's one of them, Senior fellow for emerging technologies of the German Marshall Funds, Alliance for Securing Democracy and Lizzie.

Thus far are.

The right guardrails being thought about here from a global perspective.

So I do think that the right issues and problems are being fought about. But I will caveat that by saying that the problems that we can anticipate today might not be the problems that end up causing us the

When we look at what the EU, for example, is thinking of putting in place, they want to produce risk assessments. They want to see a summarize of copyrighted material that models have been trained on. They also want it to be flagged when you're looking at AI in general, but most notably if you're looking at deep fakes.

Does that, to your point of view, go far enough?

And ultimately, are we just in a game of whack a mole all over again?

I certainly hope not. I think the EU has obviously taken one of the most aggressive approaches to thoughtful regulation of AI with the EUAI Act and the risk based approach now and the USNST has also put out a risk based framework on AI that's the National Institute of

For the models.

Now, the EU has the General Data Privacy Framework, the US doesn't have that, and it's a little bit concerning that. Despite the enthusiasm with which lawmakers are tackling the AI issue, which is a positive, that we're still not able to

It is notable, isn't it.

Every conversation we have tends to involve regulation. Everyone seems to agree it's needed and not just self regulation, but the applications of so doing is still sort of a black box here.

Yeah.

The difference our guests notifires off on this show f Lindsey, is that if you regulate the deep learning part, the training of the models, you might kill innovation. There is a difference between regulating inference or in other words, the use of artificial and technology the tool and the development of it. Where do you think we should focus our regulatory efforts.

Yeah, this is really the debate that you think you heard in the hearing, which is do we take an approach of only regulating at the point where technology meets society, ie the applications, or do we need to regulate in the model development? And I think both have some merits. Clearly, we need to regulate in the former case and focus on the actual harms that are being caused in society or that could be caused.

But that said, those.

Harms are not technology neutral. They are not applicable regardless of how you train the model. So I think that's where it gets a little bit complicated, where actually the inputs to the training do make a difference on how the harms could get propagated once technology meets society. And this idea, though that lawmakers have said today that we had the Section two thirty where we said companies just go develop and we'll give you the space and not regulate.

Do you agree with Sam Autman that the US should establish an agency to license AI development.

I would like to hear more arguments on both sides. To be honest, I think we've just gotten to the scratching the surface on whether we need a whole new agency. But the point that I really do agree with that Sam and others made is that we definitely need much more resources, whether that's with existing agencies or in fact

Should point out as well that it's not just Sam Outman, We've been talking a lot about him, but IBM's Chief Privacy and Trust Officer Christina Montgomery is also testifying in that Caroline and Lindsay Linda Gorman, Senior Fellow for Emerging Technologies at the German Marshal Fund, that thank You Grinder reporting first quarter earnings with revenue of fifty five point eight million dollars, up from forty three point five million dollars a year earlier around twenty eight percent gain year

Line or an EPs.

But what is the difference in the environment at the beginning of this year versus last year for people interacting with one another?

Yeah, so ook, I've only been on here for six months, so I can only speak about a beginning of last year per se.

But this was and no one looks backward looking.

And this year was really good and we grew really well. Our EBIDA was really strong as well, at thirty nine percent. Both growth and EBITA numbers exceeded our Follier guidance. The I think on the EPs side, there's just some stuff with warrants that makes that number look very different because of how you do work accounting. But that doesn't really impact how our business is doing, which is extremely well.

So for what's driving the engagement, what is it that people are turning to Grinder for what is the behavior you see.

Grinder has always had a lot of success in getting each new generation of people to come on the platform so on. Like a lot of other social media platforms that lose users in the new generation, we've actually not had that. We've kept users coming in and I think it frankly, the fact that all the users are there is another reason why people come back right like that, That engagement is really strong. And also Grinder's very authentic.

Let's talk about the innovation. And in some ways it feels a bit old school. George, you've got Grinder Web.

Talk to me about why you listened and heard from your user basis that's what they wanted totally.

So yeah, we just launched Grinder Web. Kind of launch it this month in anticipation of Pride next month. And the thinking is that there are a bunch of features you can launch on the web that you can't really launch anywhere else because of limitations that the app store has put on you, and that's something that our uses

Monetization one are ways in which you're thinking about people using it paying for it in a different, more seamless way on the app itself and indeed on web totally.

So what we've heard from users on kind of our subscription basis is they want two things. Number one, they want a slightly higher so a slightly lower cost model where we have an entry point that's not nineteen nine nine, but something lower than that with less features obviously because we do offer a very broad base of features in our nineteen nine nine tier, so we're working on that lower priced tier. And then they also want us to build more features on the higher end and are willing

For this past week, one Elon Musk said that Twitter or x the everything out adding some dating functionality was an interesting idea.

How seriously do you take that threat?

I mean, let's see what he does. I think it'll be interesting how he approaches things. And look, you never better against even musk rat. Nobody else has done what he's done in terms of how many companies is built. But we feel really good about our user base and the fact that our users very engaged with us. I think our engagement numbers are are frankly, very unusual. Nearly an hour of spending the app last year one hundred

The CEO, George Harrison here in San Francisco, Thanks for your time, Carol.

Yeah, let's just stick on the awning scene for a moment and look at the Chinese giant by Do shares traded here in the US.

Actually, you have been performing really well.

You see up three and a half percent there or thereabouts, and looks as though the company really managed to beat expectations. Stronger than expected revenue ten percent, it grew after its advertising in cloud businesses. This seemed to be benefiting from China's post pandemic reopening overall, so some strength there ed.

Meanwhile, coming up that we've got to talk.

About some warries post Silicon Valley bank collapse. The former CEO Greg Becker currently testifying before the US Senate committee.

They're busy up on the hill today.

He has to say, of course, about the focus the failure and what is next for regional banks more generally.

We can dig into that in a minute. This is Bloomberg.

We of course are dissecting everything that has been happening across in Washington, not only the AI focus of some outman, but Silicon Valley Bank focus Greg Becker. He's been testifying in front of the Senate Banking Committee today blaming social media in fact, for the bank's collapse. A slipe deeper into it all without one to no nationalibascond Sanni, he's gonna have a tough crowd. What ultimately do you think he's saying? Is it landing the way he hopes it too?

Do you think what is.

Interesting among the lawmakers here is you're hearing kind of two sides of a story. One is a set of lawmakers that are extraordinarily frustrated by the way SVB was managed and how much Greg Becker was paid. He declined to say whether he would give back some of his bonuses as a result of the bank's failure. And the ultimate you know, blamed mismanagement. There's a sense here also among the separate set of lawmakers that believe that a lot of this was brought on by the Fed's mismanagement

These horror It is becoming bipartisan.

It is becoming certainly very very very political on on tip of and we know also for its own worth here, we know that FED officials are also being grilled in Washington this week, and they have also outlined some series of their own failure of oversight for some of these firms as well, and.

The FDI see of course, Sinnati, we have been listening into that, Harry. Let's take a listen to just some of what greg Becca had to say.

I believe that svb's failure was brought about by a series of unprecedented events. Despite stark differences in our business models, news reports and investors wrongly lumped SVB and silver Gate together. Rumors and misconceptions quickly spread online, culminating on March ninth with the first ever social media bank run, leading to forty two billion in deposits being withdrawn from SVB in ten hours, or roughly one million dollars every second.

He claims different facts as SVB Silvergate, that did we hear any sort of mission? Shinnati quickly on him taking responsibility for what happened.

I don't think he took responsibility for all of it, and I think that's the important part here. Remember this idea of a social media fuel bank run. There's a lot still to be understood about how that happened, because remember every bank CEO across the country and every private equity from trying to back them is now having that

Clay out today, all right, bloomboch Naali Bask on the Wall Street Beat, but a story that really hit here the heart of Silicon Valley.

Welcome back to bloembog Technology. I'm Caroline hind and yelks Ed Lovelow.

In San Francisco. All about AI. But let's get a check in on the market. I think there's still a lot of lingering concern care about the debt ceiling and the progress that we're seeing or lack ofb in Washington, DC and AWS that one hundred outperformance in tech up half a percentage point.

We've been talking about do having strong earnings.

That is an outperformer in terms of the US listed shares of China tech. Ali Barber reports later in the week. But there are some movers like ten Cent to the downside, the NAS that Golden Dragon China Index, that basket of US listed China shares down half percentage point, yields climbing higher six basis points three point six percent ish on the US ten year yield where we were kind of in March, Bitcoin back down towards twenty seven thousand US dollars per token.

In terms of the individual.

Movers, I talked a little bit about bay Do that moving to the upside, strong earnings beat top and bottom line, the story about a rebound in China after the Chinese New Year easing of restrictions, travels stargning back up. Also the ad business doing well and AI AI everything all the time AI.

There's no sort of main catalyst.

But you know, Alphabet, parent of Google, has seen gained since Google Io when we learn more about it's offering. Amazon Bloomberg reporting making moves bring us some of those details, because that is what the real mover is.

To the upside is oh perfect segue ed into talking tech first up, Alphabet, you were just talking about making back market cap ground as its artificial intelligence gains. As you back on, remember it seemed to be lagging behind its peers, wasn't it, in the race to deploy generative AI products.

Abundance of caution many would call it, But Google's parent.

Company is added more than one hundred and fifteen billion dollars in market value. Some some veiling plans for its EI tools last week at the event you were at. Meanwhile, if you've been scouring the job boards for AI related jobs, Amazon I have some listed. The e commerce giant is planning to add chatchipt style search function to its online storm.

It's scorning to job posts reviewed by Blue Vegnus.

Of course, as part of a larger effort to rival efforts by Microsoft, by Google tweet these generative AI elements into its own search engine set.

All right.

Healthcare startup Hippocratic Ai staking its claim in the large language model boom with its tool, which it hopes to make medical care more accessible with The Palo Alto based company launched out of stealth today, raising fifty million dollars in a seed round from Andres and horror Itz and General Catalyst. Here in the studio with me Sinjo Munjawsha, sorry, CEO of Hippocratic AI. Fifty million dollars for a seed round. That's a lot of money. Tell me, let me just

You know, we're not announcing that today, but we really felt that generator of AI has really captured the imagination of really the world right, and it's captured the zeitgeist. And when you think about its application to healthcare, you realize, you know, there's three million missing healthcare workers in this country. We do not have enough people after the pandemic, so many burned out and there's really no way to close that gap except using technologies like Generative AI.

To do that.

But the misconception out there this morning is that you have released a chat gp T style bot that replaces the doctor the physician.

That's not the case at all, though.

Now now you know, we actually, I don't think the Generative AI is ready to do diagnoses. We think diagnoses needs to come much later when these models are safe. We think that there's a set of applications, you know, healthcare is way bigger than just the doctor. There are so many people.

Who work out back office, and not just back office.

You have registered dietitians, you have genetic counselors, you have many other roles that are supporting roles and supporting actors in the healthcare system that really could benefit from genitive AI.

I'm interested in what you see as the regulatory environment with which you put this. People are going to be fearful of medical advice plus chatchipt, many already knowing that you shouldn't. You're not going to get an abundance of advice when you go into open AI's product, when you go into bard And interestingly, we're hearing from some Oltmann at the moment thinking about regulator pressure, saying that for open ai, for Google to take on board, but they

You know?

I think that this is an area that does need a regulatory framework and one that allows us all to create safe large language models in safe generative AI. You know, we decided as we were building Hippocratic AI that we wanted to be safety first. Like this was the foundation of how we designed the company and how we designed the product. I mean, you know, our name is hippocratic, like the hippocratic ohth Our tagline is do no harm.

Share with Yeah, it's delve in because we understand your startup has the AI is possible that one hundred healthcare certifications and outperformed open AI. What is the data that you're training on and what are the safety prints you've wound into that.

Yeah, so we built in four or five things to really make this safer. I mean, I think the first was we certified it. We didn't just say, hey, let's pass the us mL the medical licensing exam, where some

Hey, it's important to be certified.

These are actually the exact exams you see at the end of your nurses name on her badge.

She has all those little letters.

We made sure we had all of those.

Those same certifications used to hire healthcare workers, or the SA certifications we used and tested on our language model. Second, we actually recruited those exact same healthcare workers. So pediatric nurses came in and gave our system feedback on how it was doing on those questions. Our dietitians came in and did the same. And so we believe who's best to judge the accuracy of a healthcare of a language model than the people doing those.

Exact jobs today.

And so.

You know, those are two of about four or five different things that we've done to really make this a safer system.

Let's do a quick far around when do you make the LM open to the public.

You know, we have decided on a threshold based launch strategy than a time based land strategy. We're saying when those professionals that I just told you about. When we have the dieticians using it and they say this is ready to go out, that's when.

It'll go out.

How do you monetize.

We will figure that out after we make sure that we build a safe and ready language model. I think again, you can't say your safety first and be like I'm launching on this date. You have to say I'm launching when the language model is ready and the professionals who do that task today say it's ready.

Minchelle, you said, and I'm sure you didn't. It was a sort of a comment that just comes out, you said. Nurses she and I'm in many ways when we say yeah, a lot on urs is our female. But therein lies some of the issue and the concern around bias within these sorts of AI models and the data that's run.

I'm sure you're.

Thinking deeply about how, particularly from a medical perspective, biases aren't built in. How do you counteract for that going forward?

You know we've already begun testing the bias of the model. You can actually go on our side at hippocraticai dot com slash benchmarks and you can take a look at our first pass of assessing the bias of our model on a bunch of different dimensions, including certain ethnic biases and certain gender biases. And you know, so far, you know, we were able to show less bias than GPT four.

But that's just the beginning.

That's our first installment, our down payment on really just trying to say, hey, look, we just launched, but we're already testing this, we already care about this, and it's something that we're going to continue to work on each and every day.

I just put out to our audience. I wrote about this this morning.

Hippocratic AI they benchmarked, were tested against Chat GPT in one hundred and fourteen certifications. I went to open ai and asked for them to comment on that performance relative to Hippocratic AI. Open ai did not reply, just putting that out there to our globalance Caroline, thanks to munjow

What is it and why.

Laski is a early It was early stage startup that does recruiting, so it matches employers with candidates very straightforward. It's not exactly clear why ex corporate Twitter was interested in buying this company, but Musk has talked about this idea of creating an everything app, so it could be a part of that broader vision.

So we're all left kind of trying to fill in the dots. Meanwhile, Lasky doesn't seem to be operating anymore online and I'm waiting for the latest tweet out of their pretty active tweeter in chief fir CEO.

Does a lot online.

What remind us of the overall vision of the X product because many would say that's why the new CEO of Twitter, for example, has come.

On board, right.

That's one thing that Musk has said when he announced that the hiring of Linda Yakarina officially on his Twitter account, he said, this is going to be part of the vision to create this everything app or this ex app, and so it must has been pretty vocal about this.

Asia. It's great to couch ut with you. Thank you, Asha Counts.

Meanwhile, we've got our VC Spotlight next Sad and it's with someone who well knows Eno Musk and is back to him and some of those other companies from New York, from San Francisco.

It's a bloomberg.

The last ten years of financial services has seen an extraordinary change. The advent of digital banking, mobile banking, and all of the technology enabled services that we have now come to consider to be table stakes. The next ten years is going to take that technology impact and expand it exponentially.

There are enablers to that that the primary driver.

Is going to be related to the deployment of artificial intelligence.

Former JP Morgan Executive Live Masters. They're speaking about the big innovations that will reshape fintech space in particular, but actually every industry when it comes to AI. Let's bring in Andrea Lamari from Manhattan Ventures Partners.

For more on the world of investing.

The world of well in fact, your own portfolio is fascinating discord of course in many ways AI driven. But Klana one of those companies of fintech business that's already used the chat GPT plug in and open AI plug in. How are you thinking about the ethical way in which these companies do adopt and let this generative AI run loose on their own proprietary data.

Thanks Carolyn, So overall we are really thinking about generative AI as a tool for good. But what it is showing to be is that companies are learning so much more about their consumer base and allowing their consumers to spend time showing trends in a much more vulnerable way,

Have if financial conditions are getting tighter, and it's hard out there for founders all bench capitals being pushed into making AI related investments that in any other economy or environment they just wouldn't normally make because of all of the hype.

Overall, what's so interesting about the vcs in the space making AI bets is that so many companies we're already utilizing AI functionality as a way to harness data and then use insights of that artificial intelligence data to build better products. But what's funny is that vcs today are having a hard time determining what's actually going to make money at so some of these companies are really taking a push towards open sourcing the technology versus actually making

I'm going to jump on that.

What's actually going to make money you invest in SpaceX? When is SpaceX going to make money? Is SpaceX ever going to IPO or spin off Starlink?

Well, overall, SpaceX does have a lot of cash and they do make money by way of the Starlink spaceships and satellite services. Because overall, what's fascinating is a lot of people across the world now that it's in every continent, are spending money on Starlink and it's generating a ton of revenue. But what we do think is that Starlink is big enough as a business that they could spin it off eventually, right, And I do see an independent IPO.

Though.

If that were the case, talk to us.

A little bit about tomorrow's IPOs today, which is what Manhattan Ventures Partner's sort of tagline is. You're all about the secondary market in many ways, and just what is the secondary market like for an E on back company.

I mean, whether it be the vision of.

X, whether it's SpaceX, so whether it's any of the companies you have in your portfolio right now and people willing and able and wanting to buy the secondary market.

So right now, it is absolutely fascinating because you're getting to see that the secondary market is driving the true value of every single private company. And overall, I think it's the real indicator for where investors are willing to buy and shareholders are looking to sell. So I think we've never seen a more clear opportunity to dive into

I mean, isn't it ed at the moment when we're ever talking about primary round and financing.

It tends to be AI related in some way shape or forward.

Does Yeah.

I think that's the part that we all want to understand better, Andrea, which is just forget a new cycle or a hype cycle. What are you doing to wake up each day and say, Okay, here's where we're going to deploy capital, here's our plan for twenty twenty three. You know, what I'm trying to understand is what is driving investment visis for vcs right now.

What's interesting about the thesis driven approach with all vcs is that all the companies that are approaching us and we are approaching them for capital raising, is that they're really looking to preserve the valuation that they last had and it is almost a valuation at all at all, stake at all, you know, permutations of evaluation. So what's interesting is that from a thesis perspective, we're determining whether or not to invest in companies that have investor sweeteners

Andrew, after Linda Yakarino news came on Twitter, did that bump the valuation you've been seeing for Twitter? I know it's part of your portfolio post it going private? Is that something that you think is getting a clearer destination, getting investors more interested.

Having Linda involved, we think is a very strong indicator of where we believe Elon is playing the sea level shuffle what we like to call right, so bringing in the strong adults in the room to really bring in a background in advertising and media, which overall that is what Twitter really is focused on, right, advertising, media and content driven.

So I think it's a strong single Linda is.

A phenomenal executive with an incredible background that I think many of us in the industry are very impressed with their ability to land. So I think generally we consider that a massive positive.

Manhattan Ventures partner Andrea Lamari, who's able Caroline to talk about every news item this week and has investments related to every sub sector that we discuss, so thank you so much for your time. In other news, East Ventures, south East Age's most active early stage tech investment firm, raise two hundred and fifty million for its twelveth fund, a rare sign of confidence in the global tech sector

Going viral is the unusual gift from Google co founder so I gave Brin. In a filing that we saw Monday, it shows that he gifted Alphabet shares worth six hundred million dollars last Thursday.

It's actually kind of.

Unclear who's got the five point two million shares. We of course know that it could be perhaps to charity or to a trust or another financial instrument. But ultimately this is on a week where Brinn and his co founder Larry Page saw their wealth are combined eighteen billion dollars and we know why.

It's because of AI.

Yeah.

Look, it all comes out of momentum from Google Io where we finally understood how Alphabet takes its competence and puts it into its tools. But I was there right and I would say they stressed deliberate, slow roll out safety guardrails.

And now we have Altman speaking on Capitol Hill.

Yeah, who's again been stressing safety.

We know that it's wrapped up now the Senate hearing that features some Oltman, CEO of Open AI, as well as IBM, as well as scientists, But ultimately he's really saying, look, there are no plans for chat GPT five training in the next six months.

This seems to be sort of under some juriss we'd heard that from some of.

His colleagues on Twitter, saying, look, whenever you might be hearing GPT five isn't currently underway. But notably, everyone wants to know what this has on the startup culture, what this has in terms of society.

Too, And again he Altman says the pressure should be on the leaders open AI and Google. The name check that was Richard Blumenthal, the Connecticut Democrat who chairs that Southern Committee, and he also had questions about the US leadership in this field.

Yeah, Visa v.

China something that we continue to discuss. How can you regulate without offsetting innovation? And what a thoroughly deep dive show we had today that does it for this edition, though on BlueBag technology. Tomorrow, well, we've got Patrick Zong Fanning partner of thirty one from Salt Eye Connections in New York.

You don't want to miss it.