voiceover: this is bbc news. we'll have the headlines and all the main news stories for you at the top of the hour, straight after this programme. this week, paul has

been tackling blazes in germany. question, how do you attract the right fish to your net? we will shine the light on a colourful solution. oh, look at that, pennies from heaven. and... take a feel. you can feel the texture. it's really soft. yeah. that's not what you need from a crisp. we'll find out why it's crunch time for al. it is an inferno and it's leaving little in its wake. we've seen wildfires raging across many countries over the summer months. the hot, dry conditions exacerbated by climate change are making them more frequent and more intense. and it can often be a real struggle to bring them under control. high—tech solutions are changing how we tackle them, whether it's drones

dousing the fires or satellites predicting their spread. and paul has been to a forest near berlin to find out how one company is trying to prevent forest fires from taking hold in the first place. it's conventional wisdom that the earlier forest fire can be stopped the better, but often a fire can already be well established before it is noticed, making it harder to put out. but one company in germany believes it has a possible solution by using internet of things sensors and ai to understand what forest fires smell like. the basic function is like an electronic nose that we put into the forest and it can actually smell a fire. we can detect fires during the smoldering phase. so a few minutes after someone has thrown a cigarette, we should be able to ring the alarm bells.

every forest fire has a unique aroma based on what surface material is around. i love that smell. yes. their forest ecology expert, doctor juergen mueller, collects these samples from around the world and then burns them to generate the data to train the ai on what an actual fire in a specific location may smell like. ok, now, we close the chamber. this test, we have 400 degree. and so the smoke�*s being drawn into this chamber? you can actually see it moving across into the air. yeah. so it doesn't need very much smoke. i thought they might be like this... this is enough. this is our fantastic idea. we need small smoke. here we have the sensors. and the sensors in here will be

be able to detect when that smoke comes through, but also be able to tell you what sort of smoke that is. right. these tests allow the team to build a vast bank of data and help spot real fires and avoid false alarms from things like cigarettes or exhaust emissions. but in lab conditions is one thing, but what about out in the wild? it was time to start our own forest fire. under strict supervision of course. we can see is it a realfire or is it something else? the main cause of uncontrolled wildfires is us humans. it's often barbecues, cigarettes or glass bottles which start a fire. because of this, dryad looks to put the sensors every 100 metres in more remote areas and 15 metres

near well walked paths. so it looks like quite a lot of smoke coming off there. would that be enough already for your sensors to be able to detect? that is enough for our sensors. so this smoke needs to reach the sensors. that means if the sensor is close to the smoke, for example, on this tree, then you have it within seconds. yeah. but it's notjust the ai and the sensors that are innovative, there's another problem that needed solving. we are here in a forest, in a rural area where i imagine there isn't much mobile signal. how do you get the data from the sensors out to where it needs to go to? because mainly there is no internal telecommunication infrastructure in the forest, we bring our own telecommunication infrastructure to the forest. so this is the mesh gate. oh, wow. 0k. yeah. these mesh gateways need

to hear over the air if one of these sensors in the forest has an alarm or a signal and so on, and then it receives the signal from the sensor and sends it to the next mesh gateway. it is like a relay to hop as a relay. it's a border. we have the so—named border gateway to connect to the internet. so is there a limit to the area the size that you could that you can use this system in? or is it simplyjust the number of relays you could go as big or as wide as you want? we can in theory, we can go as big as we want. so we have a range of approximately two kilometres. and then let's say every two kilometres you have here a mesh gateway. and then you can from hop to hop to hop. who are your customers or your potential customers, your target market for this? well, our customers, we currently have about 50 deployments in southern europe and north america predominantly are private forestry.

to some extent, that needs to protect the forest and the stands that they have the value. but then municipalities that need to protect the areas and the people living in proximity of forest and utilities, in particular power line companies and in railroad organisations that actually face the problem that their products, their services may cause wildfires and they might be liable for the consequential damages. and we have them to protect against. they believe that this network could ultimately support wider forestry monitoring. what we want to do with our networks is actually to become the iot backbone of the forest. so we want to help, to protect, but also to restore nature. and that means we will branch out from fire detection, which we do today, to soil moisture monitoring, tree growth monitoring, detecting chainsaws to prevent from illegal logging. so there are all sorts of applications that you can

build with iot, once you have a telecommunications network in the forest. but dryad is one of many solutions trying to tackle the wildfire problem. and it's still early days for this technology. now, as artificial intelligence is rolled in, our homes, workplaces and even social lives continues to grow you'd be hard pressed to find an industry that's not using it. but al's fuel is data, and data requires storage, and that uses a whole lot of energy. yeah. and a new study from fryer university in amsterdam is warning the technology could consume as much energy as a country the size of the netherlands by 2027. zoe kleinman has been taking a look. ai is everywhere, from chat bots to facial recognition software, virtual assistants to disease diagnosis. the technology is rapidly becoming part of 21st century life. what drives ai tools are huge

networks of computers stored in massive data centres like this one here in scotland. the more powerful ai gets, the more mighty machines it uses. and they need energy and they need to be kept cool. a new peer reviewed study has warned the ai sector could use the same amount of energy as the netherlands injust four years' time. the key lesson should be that using ai can make an application more energy intensive. and just for that reason, you want to be very mindful about whether you want to use ai in the first place. alex says his research is intended as a warning rather than a prediction. experts say they can only speculate on figures because much of the data they need is shrouded in secrecy. the tech firms don't share enough of it. but he's not alone in sounding alarm bells about the environmental impact of the tech. no one knows the price at all. a standard rack full of normal kit is about four kilowatts

of power, which is equivalent to a family house, whereas an ai rack would be about 20 times that. so but 18 kilowatts of power. is there more that you could do? we use a technology called free cooling as a massive overhead for datacentres and it consumes a lot of power. so we use the scottish climate. we take the cold air from the outside and we mix that into the hot air that the equipment generates. a traditional data centre in the past would have used refrigerant and cooling systems that generate a lot of power, so the free air cooling can reduce that by up to 80%. ai is going to use a lot of power and generate a lot of carbon, but there are small steps that they can take, for instance, making sure that these ai workloads are deployed in regions of low carbon intensity. so use green energy like scotland, for instance. but is there anything ai could do to help solve the problem? well, i can definitely be of some use.

i mean, in the end, whenever you're building these models, you're going to be left with something that can do something. and it depends on what kind of contents you're using these models for. the most well known models right now are language models that are used in chatgpt and similar services. but of course, you can also use it in all kinds of ways. you can use it for climate modelling. google and american airlines recently found pilots could halve the amount of contrails created by planes using al to select their altitude. contrails, those streaks in the sky contribute to global warming, and the us government is spending millions of dollars trying to recreate nuclear fusion — the way the sun gets its energy. the hope is that al can help. success here would give us a limitless green power supply, but it means we have to gamble that the environmental price tag of building al to find out is one the planet can afford.

i'm alister keane with a look at this week's tech news. the winners of the 2023 earthshot prize have been announced with five companies winning £1 million each. prince william presented the awards in singapore. it's the third year of the awards for companies tackling climate change have taken place. a former senior staff member at meta says instagram is not doing enough to protect teens from sexual harassment. meta says it's brought in over 30 tools to support a safe environment for teens online. but the former worker says he realised there was a problem when he saw his daughter's experience on the social network. i was shocked when i asked her, so what did you do? and she said, well, there was nothing i could do because i had no option to report it. so then ijust blocked the person and they asked her and then what happens? and they're like, yeah, you know, theyjust laugh it off and move it on to the next person. the founder and ceo of bumble, whitney wolfe herd, is stepping down after nearly

ten years in charge. she designed the dating app in 2014 to be a space where women make the first move. she'll pass the baton to slack boss libby—annnjones. and finally, nintendo and sony have announced a live action version of the gaming classic the legend of zelda. there's no word yet on the plot or cast, but it will be co—produced by video game designers and film makers behind the spider—man trilogy. now the fishing industry is big business, but one of its big problems is something called bycatch. now, bycatch describes fish and other animals such as dolphins and sea turtles that get inadvertently hooked or entangled in fishing gear. this bycatch is then usually discarded, overboard, either dead or dying. environmental charities have been working with the fishing industry to try to reduce the problem.

but one fisherman in scotland hopes that technology can help. my name is ian wightman. i'm a langoustine fisherman on the west coast of scotland. i started fishing when i was 16, and i've been a skipper for 39 years. well, i've recently become involved with a company who have developed a camera we can put onto the net and it gives us footage that we never had any access to. led lights, pisces lights, another piece of technology that is called end key, which is a data collection system. these are all things that we can use in our arsenal to try and catch more efficiently.

usual monday morning routine. what we have here about the pisces lights such as onto the net, you can have them flashing. we've got, i think we've got five different colours which will either attract or repel fish depending on the colours and the configuration of them. so we're now learning the behaviour of fish to the different colours of light, because some light does one thing to one fish and does another thing to another fish. we have the cash count, which is an underwater recording unit which will give us visibility on the net and the water, which is something that we've never had access to. the cameras are extremely easy to use that in purpose made cartridge you latch onto the net. in the olden days, everything

was done by expedience. you would make adjustments and you could spend maybe all day with your net on one adjustment. you come out the next day, change it again, another two or three hauls, so it could take you three or four days to make your net fish what you think is the most efficient way. whereas with catch cam you can see exactly what your gear is doing. so it takes the guesswork out of it for us. so this is a key sensor and the benefit with the key is to put these end dates and record data. you're measuring temperature in the water, which we use in fishing. when you get the growth in the water. you'll get better prawns and shallower water to target better quality prawns

without the guesswork. i pull the strings and the stuff will come cascading down. oh, look at that! pennies from heaven. so part of the theory behind the pisces is for not to repel fish, but to attract fish like the squid using the the lights on the squid nets to help attract the squid in towards the gear.

that helps improve the catches, improves the efficiency. you can see how the catch and the fish that has been eaten by seal. well dead. sure we got one, two small fish. put these units on six or seven boats, probably for the cost it would take for a week's charter for a government boat. so the potential for gaining data is phenomenal. and as a fisherman, we have to be involved in the science. do you mind? sorry. just making the point. what is the point? crisps are noisy. well, yes, i can hear that. yeah, but it's notjust the eating of crisps. the process of making them is noisy too.

too loud. factory. exactly. but did you know that as the machines in the factories wear out, they make ever so slightly different noises? and if you know what you're listening for, you can keep everything running smoothly. 0h. no thanks. here's alister keane. this factory is producing snacks eaten the world over. take a feel. you can feel the texture. it's really soft. yeah. it's a fast business with thousands of bags of snacks leaving the factory each day straight to the supermarket shelves. keeping all these machines going is vital. in the process area, they are critical. when they stop, the whole line stops. and that's what we focus

on is where do we stop our, you know, the beeper if we look at the big curve, what stops ourfryers, ovens, etc? to try and predict when the machines aren't right before they break down the factory�*s owners, pepsico, are turning to ai—powered sensors, but the secret ingredient is something the factory isn't short of. all of these machines are really noisy. yes. they're listening to the sound coming from each bit of machinery and analysing the health of the equipment. there is up to four sensors on every single asset, so you start to multiply that out. we've got a lot of sensors on our equipment. and how do you find using them and using the reports they give off? it's really a game changer for us. you know, pepsico are starting to get into that digitalisation world. and, you know, this is a major part of that for us. we're starting to change how we're working. we're starting to

integrate the data. we're making decisions through data analytics. yeah, it's really taking us to a different level now. away from the noise of the factory, we can explore how it works. at agri, we work with the largest manufacturing companies like pepsi to make their production lines and the machines they operate more reliable, more productive and more sustainable. and we do this by basically listening to the machine. and based on the noise, we can tell you what's wrong with them. so we have our own sensor that we put on the machines and we measure three things. we measure vibration, temperature and magnetic emissions from the motor. and by analysing this, we can detect malfunctions months or weeks ahead of when they happen. so this is what a normal machine sounds like... and this machine has a malfunction. let's see if you can pick up what it is... so it's much more high pitched.

but beyond that, i would have no clue what's wrong. so this is where kind of a sophisticated ai algorithm can really detect the specific issue that the machine has. vibration and sound are physically the same thing. right? and we have today over 300 million hours of machines that we've we've analysed and monitored. and we can leverage all of this database to create algorithms that know how to pinpoint the specific patterns of different malfunctions. so if you have, let's call it a bearing wear versus another type of fault, we can detect exactly what it is and tell you what you need to do to fix it. detecting the problems in this way means repairs can be scheduled and problems fixed without causing a shortage of products in the supermarket. on top of it, we talk about sustainability, right? so if the machine is working in the most optimal way, you can reduce the energy consumption of that machine. you can reduce the waste that the production line

manufactures by improving the quality of the product. this tech is now being rolled out across the compa ny�*s global factories after a pilot in the usa proved successful. our team was set up to work hand in hand with the business, to understand what are the real—world business needs. we then scout around the world to find the most cutting—edge technologies. our goal is to rapidly pilot the technologies to make rapid go/no—go decisions. and then if it works, if the solution really proves itself, scale, help the business scale a solution around the world. what we liked about algri is that it brings to technologies together its sensors that we could track what's happening in ourfactory, we could listen to our factory, understand sound, temperature, vibrations. but also we're able to apply ai and artificial intelligence to take all that data, pull it together and get actionable insight. if your factories, though, are being more efficient

and producing more products, does that mean you could close some factories? people are at the centre of everything we do at pepsico, we see solutions like algri helping make our lines more efficient. it's helping enhance how people work so we can bring better efficiency to meet the needs of our people, of our customers, and we can be prepared into the future to meet the needs on a daily basis. tech in factories is constantly evolving, but in the case of these snacks, i'm ok with the taste staying the same. that was alastair. i'm afraid that's all we've got time for. yeah. thanks for watching. we'll see you soon. bye.

hello there. we're starting our sunday morning off on a cold and frosty note across much of the north and east of the uk, with some lingering dense mist and fog patches around. but a big change taking place across more southern and western areas. we've got a weather front spreading northwards, that's going to bring outbreaks of rain and we'll see the rain and cloud spill northwards through the course of the day. i think it's much of northern scotland which will tend to stay dry with some sunshine all day. now some of the rain will be fairly heavy in places, particularly in towards northern ireland. the breeze picking up as well, but cold, frosty and foggy with some sunshine across northern and eastern areas. there could be just a few showers peppering northeastern coasts of scotland in towards the northern hours, but most places will see the sunshine, areas that don't have mist and fog, mind you. but northern ireland into wales, the midlands, southern and western parts of england will start cloudy and wet and it looks pretty damp here for the remembrance sunday

services. now that area of cloud and rain will continue to move its way northwards. we'll see a little bit of brightness pushing into the southwest and south wales later on, but even here we'll have a few hefty showers. temperature—wise turning milder in the southwest but cold across northern and eastern areas. and like i mentioned, it's northern and eastern scotland which will hold on to the sunshine. through sunday night, that weather front eventually pushes northwards. so with cloudier skies across scotland, northern england, it will be less cold, but it turns much wetter and windier across england, wales and northern ireland towards the end of the night. and very mild here, 12 to iii degrees, but less cold across the north and i don't think we should see any frost problems. this renewed area of low pressure could cause some issues. we've already got a yellow rain warning in force for northern ireland. the rain will be heavy and persistent and we could see gales, even severe gales, around the irish sea coast as we move through the course of monday. that rain pushes its way northwards into scotland through the afternoon. it brightens up for northern

ireland and certainly for england and wales later on, could see one or two showers around, but it will be the temperatures you'll notice on monday, much milder, 12 to 15 degrees, even 8 to 11 degrees across scotland. it does remain fairly unsettled, though, as we move through the new week with areas of low pressure bringing more unwelcome rainfall, but the signs of it perhaps turning a little bit drier and brighter for many areas by the end of the week.

live from london, this is bbc news. israel agrees to help evacuate babies from gaza's al—shifa hospital. iceland declares a state of emergency after a series of earthquakes raise fears of a volcanic eruption. uk prime minister rishi sunak condemns protests seen across london yesterday as "violent and wholly unacceptable." police say 126 arrests have been made. the israeli military says it has agreed to help evacuate

babies from gaza's main hospital, al shifa, where conditions have become desperate while fighting